Vollna AI Bot for Upwork Projects

Your hashtags (#GP, #Visitor, #LCI, #Vicki, #Sphere, #Wedding) are already acting as segment identifiers inside BoldTrail. That's actually perfect, it means the foundation is already there!

Here's what I'll do with them:

#Sphere / #Vicki tags — these are people who know you personally.

I'll attach a warm "Stay in Touch" drip campaign to this group , friendly check-ins, market updates, personal touches

#GP / #LCI / #Visitor tags — these are colder leads. I'll attach a Buyer Nurture sequence to these, more educational content, listing alerts, calls to action

Assign campaigns directly TO a hashtag group, so every contact wearing that tag automatically gets enrolled in the right campaign. You don't have to touch each person one by one.

with the birthday and age data you uploaded, I can also set up automated birthday messages and target the right people for downsizing vs upgrading conversations, that's powerful stuff most agents never use!
#Wedding A warm congratulations email with their wedding/reception photos and a heartfelt personal note from you
A gentle follow-up a few weeks later - "Now that you're newlyweds, have you thought about finding your first home together?"
A third touch with a simple first-time homebuyer guide or

Actualmente estamos buscando una persona que pueda ayudarnos a mantener, optimizar y seguir desarrollando nuestro ecosistema de automatización para generación y seguimiento de leads dentro de nuestro negocio de bienes raíces.

Hoy en día ya contamos con un sistema funcionando y buscamos alguien con experiencia real trabajando sobre automatizaciones complejas, integraciones y lógica de funnels.

Nuestro ecosistema actualmente incluye:
Meta Ads
Landing Pages
ActiveCampaign (CRM + automatizaciones + pipelines)
Twilio
Integraciones vía API
Bot conectado con OpenAI

El sistema ya está operativo, pero queremos seguir mejorándolo, creando nuevas automatizaciones y optimizando procesos internos.

No estamos buscando a alguien para empezar desde cero, sino una persona que pueda integrarse a una infraestructura ya existente, entender cómo funciona y ayudarnos a evolucionarla.

Primera tarea / proyecto inicial

Queremos comenzar con una implementación específica como primera prueba de trabajo juntos.

Actualmente, cuando un lead agenda una llamada con nuestro equipo de ventas, necesitamos automatizar una secuencia de recordatorios para mejorar la asistencia a las reuniones.

La lógica debe tomar la información desde ActiveCampaign y ejecutar recordatorios automáticos:

24 horas antes de la llamada
2 horas antes
15 minutos antes

Los recordatorios deben salir tanto por email como por nuestro sistema de mensajería automatizada conectado al ecosistema actual.

Más allá del envío de mensajes, nos importa que toda la lógica quede correctamente estructurada, sincronizada y escalable dentro de ActiveCampaign y las automatizaciones existentes.

Lo que valoramos

Buscamos alguien que:

tenga experiencia real con automatizaciones complejas,
entienda APIs y webhooks,
pueda trabajar sobre sistemas ya existentes,
tenga experiencia con ActiveCampaign y Twilio,
entienda lógica de funnels y seguimiento de leads,
tenga buena capacidad de organización técnica,
y pueda comunicarse en español.
Importante

Nosotros proporcionaríamos accesos, contexto y documentación para facilitar el entendimiento del sistema actual.

Inicialmente comenzaríamos por proyecto, pero si existe buena compatibilidad, esto podría convertirse en una relación de trabajo continua con futuras implementaciones y mejoras.

Si entiendes que este proyecto encaja con tu experiencia, nos gustaría conocer:

Tu experiencia trabajando con automatizaciones similares.
Cómo abordarías técnicamente este tipo de implementación.
Qué información necesitarías para evaluarlo correctamente.
Y un estimado de presupuesto aproximado para desarrollarlo.

We are an early-stage startup building AI-first approach from day one.

Our Phase 1 objective is straightforward:
Design and establish an AI-enabled operating foundation for founders covering collaboration, communication, knowledge management, automation, and decision orchestration.

We are looking for someone who can help architect and solution the foundational setup, tools, workflows, and operating approach needed to run lean, scalable, and AI-native.

Key Focus Areas
Communication & Team Collaboration
Meetings, Action Tracking & Decision Capture
Shared Knowledge Management & Idea Validation
Document, Content & Information Lifecycle Management
Founder Workflow Automation & AI Assistants
AI-Driven Operating Processes and Integrations

What We Are Looking For
Experience designing AI-enabled workflows and collaboration environments
Ability to recommend practical architecture, tools, and governance approaches
Experience with AI agents, automation platforms, knowledge systems, and enterprise collaboration tooling

Initial engagement will focus on:
Discovery discussions
Potential for longer-term collaboration as we scale.

If this aligns with your experience and interests, let’s connect for an initial discussion

Project Scope Understanding
Phase 1: The immediate goal is to enhance the generator so it produces realistic, QuickBooks Online (QBO)–importable CSV files.

Phase 2: This will involve OAuth integration with QuickBooks, data injection/tagging via the Intuit API, and reset-to-baseline functionality. I understand this is a separate phase, and I will ensure that Phase 1 is structured to integrate smoothly into it.

Proposed Approach
The current implementation is built as a single large file, which limits scalability and realism. I propose restructuring it into a modular engine, where realism is a core principle.

1. Industry Profiles (YAML-based)
Each industry (e.g., restaurant, retail, etc.) will have its own configurable YAML file containing vendors, products, pricing ranges, and transaction timing. This allows easy tuning without requiring code changes.

2. Shared Ledger System
All accounts (bank, credit card, savings) will feed into a single running ledger, ensuring transactions reconcile properly and the financial “story” remains consistent.

3. Realism Layer
Introduce:
Non-rounded, natural-looking transaction amounts
Authentic bank-feed style descriptions
Recurring transaction patterns (rent, payroll, subscriptions)

4. Balance Guard Mechanism
A safeguard to maintain realistic account balances by automatically introducing adjustments (e.g., owner deposits or expenses) when needed.

5. Scenario Engine
A configurable layer to simulate real-world irregularities such as:

Refunds
Duplicate deposits
Personal expenses
Bank fees
This can be dialed up or down depending on the dataset requirements.

6. Seed Control + Testing
Deterministic outputs using seed values (same input → same output)
Automated tests to validate accounting integrity (e.g., reconciliations, no invalid balances)
Suggested First Step (Low-Risk Approach)

Rather than implementing the full refactor upfront, I recommend starting with one industry end-to-end.

This will:
Provide a realistic sample for you to evaluate
Allow us to fine-tune the logic and realism together
Make scaling to other industries significantly faster (mostly configuration-driven)

currently seeking a qualified specialist to undertake the configuration of a laptop for the use of Claude Desktop and Cowork. This role entails the comprehensive setup of a laptop tailored for artificial intelligence tool utilization, ensuring that all requisite software and configurations are effectively implemented.

The ideal candidate will possess demonstrable experience in network administration and will be adept at troubleshooting any issues that may emerge throughout the setup process.

• Describe your recent experience with similar projects
• How do you operationalize your projects?

Hello,

We are building a WhatsApp automation / AI chatbot system for car dealerships in the Netherlands.

The goal is to help dealerships:

* respond faster to customers,
* automate customer communication,
* qualify leads,
* and book appointments faster through WhatsApp.

The AI should be able to have natural conversations with customers and assist with things like:

* car inquiries,
* test drive requests,
* service appointments,
* availability questions,
* and follow-ups.

Important:
In the Netherlands, dealerships use DMS systems (Dealer Management Systems), so we need integration capabilities with DMS software.

The chatbot should eventually be able to:

* read/write appointment data,
* access customer information,
* and sync with dealership workflows through DMS APIs if available.

We are looking for someone experienced with:

* WhatsApp Business API,
* OpenAI / AI integrations,
* backend development,
* API integrations,
* automation workflows,
* and scalable chatbot systems.

Please let us know:

1. What experience you have with similar projects
2. Which tech stack you recommend
3. How you would structure the architecture
4. If you have experience with CRM/DMS integration

Also Someone for long term collaboration and fast reply

Hi everyone 👋

I’m looking for an experienced AI/ML developer to join an ongoing crypto trading project and help continue development after my previous developer stepped away from the project.

This is a long-term opportunity for someone who genuinely enjoys working on:

AI + trading systems
Crypto market prediction
Feature engineering
Multi-timeframe analysis
Trading bots & automation

The project is already partially built, and I’m now looking for the right person to improve, optimize, and help take it live.

🚀 About the Project

We are building an AI-powered crypto trading system focused mainly on:

BTC/USDT
SOL/USDT
LINK/USDT

The system is designed to:

Detect LONG and SHORT opportunities
Identify market trends
Detect choppy vs trending conditions
Improve entries and exits
Support live trading execution

We already have:

Historical market datasets
Tick-level data
Existing models
Feature engineering pipelines
Telegram bot integration
Partial backtesting/simulation systems

We also already have a strong trading strategy structure that has shown very high-potential results, and we are now looking for the right developer to help properly optimize, validate, and scale it further toward our target of achieving an 85%+ success rate system consistently.

🔍 Looking For Someone With Experience In
AI / Machine Learning
XGBoost
LightGBM
LSTM / BiLSTM
Time-series forecasting
Classification models
Model validation & optimization
Trading Knowledge
Market structure
Momentum/trend analysis
Risk management
Multi-timeframe systems
Entry/exit logic
Python Stack
Python
Pandas / NumPy
Scikit-learn
TensorFlow or PyTorch
Binance API experience

Bonus if you have experience with:

Tick-level data
Order flow / CVD
Crypto futures systems
Live bot deployment
📌 What You’ll Help With
Reviewing and improving current models
Improving signal quality and accuracy
Enhancing feature engineering
Improving backtesting/simulation logic
Improving trading execution logic
Helping optimize the strategy toward high win-rate consistency
Working toward live deployment
💡 Important

I’m looking for someone who:

Communicates clearly
Can think independently
Understands real trading systems
Is honest about results
Can work long-term if things go well

This is NOT just a simple indicator bot project — there’s already a solid foundation and a lot of work has gone into the data/features side.

📩 When Applying

Please share:

Your experience with trading systems
Relevant AI/ML projects
Binance API experience
Experience with time-series or trading models
GitHub/portfolio if available
Your preferred tech stack

Would love to work with someone passionate about building something serious long-term 🙌

Will need a talented full-stack developer to assist taking a concept to MVP for an internal project.

I am 100% fine with vibe-coded deliverables as long as the developer can handle debugging and problems as they arise.

The project summary is to add a resume inside the app and then have it customize the resume to a job posting and output a specific PDF for one job.

# Developer-Oriented Summary — AI Financial Assistant for Local Governments

## 1. Product Objective

The project is to build a sovereign, secure, vertical AI financial assistant for local governments, including municipalities, inter-municipal bodies, and other public local authorities.

The goal is not to replace existing public finance software. The product must act as an intelligence layer above existing systems. It should allow users to import or connect financial data, structure it, query it in natural language, detect anomalies, generate dashboards, and produce usable reports or decision notes.

The assistant must help finance teams, chief administrative officers, elected officials, service managers, management controllers, consultants, and outsourced finance directors understand and exploit financial data more easily.

The product must provide reliable, sourced, explainable, and actionable answers. Every figure must be traceable to a source. Every calculation must be explainable. If the data are missing, incomplete, contradictory, or outside the user’s access rights, the assistant must refuse to answer or clearly state its limits.

The solution should start with a prototype based on Excel/CSV imports, then evolve into a web MVP, then into a scalable SaaS or private deployment architecture with connectors, APIs, advanced dashboards, RAG, and reporting features.

---

## 2. Core User Problem

Local governments already have a lot of financial data, but this data is often scattered across several tools and file formats. Finance teams frequently need to export Excel files, clean data, reconcile tables, manually check inconsistencies, and produce notes under time pressure.

The main operational problems to solve are:

1. difficulty obtaining a consolidated financial view quickly;
2. dependence on Excel exports and manual reprocessing;
3. difficulty explaining differences between voted budget, opened credits, committed amounts, mandated amounts, paid amounts, and available credits;
4. lack of early warning alerts on budget risks;
5. time-consuming production of notes for executives and elected officials;
6. poor readability of financial data for service managers and non-financial users;
7. heterogeneous software and data formats;
8. weak internal control tooling.

The product must therefore provide a practical interface between raw financial data and business decision-making.

---

## 3. Product Positioning for Development

The product must be designed as a vertical business assistant, not a generic chatbot.

A generic chatbot can generate plausible text. This product must:

1. use real internal data;
2. cite the sources used;
3. explain calculations;
4. distinguish facts, analysis, and recommendations;
5. refuse unsupported answers;
6. apply local public finance business rules;
7. respect user permissions;
8. isolate each local authority’s data;
9. generate usable deliverables;
10. maintain logs and audit trails.

For developers, this means the application must not be built around the LLM alone. The core system must include:

* a structured data model;
* an ingestion and mapping engine;
* a rules engine;
* a document indexing and retrieval system;
* a controlled AI orchestration layer;
* dashboards;
* export generation;
* user and permission management;
* logging and observability.

The LLM should be treated as one component within a controlled architecture.

---

## 4. Priority MVP Scope

The MVP must demonstrate business value quickly, with a limited and realistic scope.

### MVP must include:

* Excel and CSV import;
* secure file upload;
* semi-automatic column mapping;
* basic data quality checks;
* data normalization;
* budget, commitments, mandates, and suppliers as core datasets;
* business AI chat;
* sourced answers;
* calculation explanations;
* refusal when data are missing;
* simple anomaly alerts;
* dashboards for budget execution, commitments, mandates, and suppliers;
* note generation;
* Excel, Word, and PDF exports;
* basic user management;
* import logs;
* conversation logs;
* strict separation by local authority.

### MVP should exclude:

* full API integration with all software vendors;
* real-time synchronization;
* advanced financial forecasting;
* official regulatory document production;
* automatic correction in source systems;
* automated accounting entries;
* direct DGFiP access;
* certified scoring;
* replacement of human expertise.

The first product version should focus on **budget execution, commitments, mandates, suppliers, simple alerts, and note generation**.

---

## 5. Main User Profiles to Support

The system should support multiple user profiles, but the MVP can prioritize finance directors, chief administrative officers, finance agents, and service managers.

### Finance Director / Finance Manager

Needs advanced financial analysis, dashboards, alerts, reliable calculations, and fast note generation.

Typical questions:

* What is the current budget execution rate?
* Which chapters are drifting?
* Which commitments remain open?
* Which credits may be insufficient by year-end?
* Which suppliers concentrate anomalies?

Main development requirements:

* high data accuracy;
* correct distinction between committed, mandated, paid, and available amounts;
* source traceability;
* exportable analysis.

### Chief Administrative Officer

Needs executive summaries, risk prioritization, decision notes, and financial trajectory visibility.

Main development requirements:

* synthetic view;
* risk scoring or risk ranking;
* clear recommendations with limits;
* separation between confirmed data and interpretation.

### Elected Official

Needs readable, non-technical explanations.

Main development requirements:

* pedagogical reformulation;
* simplified notes;
* charts and key figures;
* no excessive technical vocabulary.

### Service Manager

Needs to track budget availability and commitments within their own scope.

Main development requirements:

* service-level dashboards;
* strict access rights;
* available credit calculation;
* alerts on overconsumption or open commitments.

### Finance Agent

Needs operational controls and anomaly detection.

Main development requirements:

* filtering;
* duplicate detection;
* old commitments;
* rejected mandates;
* imputation inconsistencies;
* exportable control lists.

### External Consultant / Outsourced Finance Director

Needs multi-client diagnostic capabilities.

Main development requirements:

* strict tenant isolation;
* report generation;
* reusable diagnostic templates;
* data source tracking.

---

## 6. Core Use Cases to Implement First

The MVP should not try to implement all use cases immediately. The first technical version should focus on a limited set of high-value use cases.

### Budget execution

The system should answer:

* What is the operating budget execution rate?
* What is the investment budget execution rate?
* Which chapters show abnormal consumption?
* Which articles increased the most compared with the previous year?
* Which credits are at risk of being insufficient before year-end?
* Which credits are under-consumed?
* What are the differences between voted, opened, committed, mandated, paid, and available amounts?
* Can the system generate a budget execution note?

### Commitments

The system should detect:

* outstanding commitments;
* commitments older than six months;
* commitments linked to nearly exhausted credits;
* possible duplicate commitments;
* commitments without an associated public contract;
* services concentrating open commitments.

### Mandates

The system should analyze:

* rejected mandates;
* recurring rejection reasons;
* suppliers concentrating rejections;
* late mandates;
* mandates above a defined threshold;
* possible imputation anomalies.

### Suppliers

The system should identify:

* supplier concentration;
* duplicate third-party records;
* suppliers with abnormal spending volume;
* suppliers linked to anomalies;
* suppliers without contract reference where expected.

### Internal control

The system should provide:

* priority anomalies;
* control lists;
* risk labels;
* simple recommendations;
* exportable action plans.

---

## 7. Data Ingestion Module

The ingestion module is one of the most important parts of the product.

It must support:

* Excel import;
* CSV import;
* secure file upload;
* future API or connector ingestion;
* automatic or assisted format recognition;
* column mapping;
* required-column detection;
* data type detection;
* consistency checks;
* import history;
* error logs;
* versioning.

The system must always preserve the raw imported file. It must never overwrite source data without history. Every transformed dataset must remain linked to the original import.

### Required ingestion behavior

When a user uploads a file, the system should:

1. store the raw file securely;
2. detect file type;
3. detect sheet names if Excel;
4. detect column names;
5. propose mapping to the internal data model;
6. identify missing mandatory columns;
7. detect suspicious values;
8. normalize formats;
9. store clean data in structured tables;
10. generate an import report;
11. log the operation.

### Import report should include:

* import date;
* user;
* local authority;
* file name;
* detected data type;
* number of lines;
* mapped columns;
* unmapped columns;
* missing mandatory columns;
* invalid values;
* warnings;
* quality score;
* status.

---

## 8. Core Financial Data Model

The data engine must structure local public finance information.

The model must support at least:

* local authority;
* fiscal year;
* budget;
* budget type;
* operating section;
* investment section;
* chapter;
* article;
* function;
* service;
* operation;
* voted amount;
* opened credits;
* committed amount;
* mandated amount;
* paid amount;
* available amount;
* commitment;
* mandate;
* revenue title;
* supplier;
* third party;
* public contract;
* imputation;
* rejection;
* supporting document;
* source file;
* import batch.

### Essential budget concepts

The model must clearly distinguish:

* voted budget;
* opened credits;
* committed amount;
* mandated amount;
* paid amount;
* available credits.

This distinction is critical. The system must not confuse these values.

### Recommended core entities

#### LocalAuthority

Represents a municipality, inter-municipal structure, or other public entity.

Fields:

* id;
* name;
* SIRET if available;
* type;
* tenant identifier;
* created_at;
* updated_at.

#### FiscalYear

Fields:

* id;
* local_authority_id;
* year;
* status;
* start_date;
* end_date.

#### BudgetLine

Fields:

* id;
* local_authority_id;
* fiscal_year_id;
* budget_name;
* section;
* chapter;
* article;
* function;
* service;
* operation;
* voted_amount;
* opened_credits;
* committed_amount;
* mandated_amount;
* paid_amount;
* available_amount;
* source_id.

#### Commitment

Fields:

* id;
* local_authority_id;
* fiscal_year_id;
* commitment_number;
* date;
* supplier_id;
* service;
* chapter;
* article;
* object;
* committed_amount;
* mandated_amount;
* remaining_amount;
* contract_id;
* operation;
* status;
* source_id.

#### Mandate

Fields:

* id;
* local_authority_id;
* fiscal_year_id;
* mandate_number;
* date;
* supplier_id;
* amount;
* chapter;
* article;
* service;
* imputation;
* status;
* rejection_reason;
* source_id.

#### Supplier

Fields:

* id;
* local_authority_id;
* name;
* SIRET;
* internal_reference;
* normalized_name;
* duplicate_group_id;
* source_id.

#### PublicContract

Fields:

* id;
* local_authority_id;
* contract_number;
* title;
* supplier_id;
* start_date;
* end_date;
* maximum_amount;
* consumed_amount;
* status;
* source_id.

#### Alert

Fields:

* id;
* local_authority_id;
* fiscal_year_id;
* alert_type;
* severity;
* rule_id;
* entity_type;
* entity_id;
* title;
* explanation;
* calculation_details;
* source_references;
* recommendation;
* status;
* created_at;
* resolved_at.

#### Source

Fields:

* id;
* local_authority_id;
* import_batch_id;
* file_name;
* file_type;
* sheet_name;
* row_number;
* original_value;
* normalized_value;
* created_at.

---

## 9. Business Rules Engine

The business rules engine must detect financial risks, anomalies, and control points.

The MVP should include simple, explainable rules before adding advanced scoring.

### Priority rules

The rules engine should detect:

1. insufficient credits;
2. abnormal consumption;
3. under-consumption;
4. old outstanding commitments;
5. commitments not fully mandated;
6. commitments with no public contract reference;
7. rejected mandates;
8. suppliers concentrating large spending;
9. supplier dependency;
10. duplicate suppliers;
11. imputation inconsistencies;
12. late mandates;
13. old revenue titles if data exist;
14. general expenditure drift;
15. payroll drift if data exist.

### Required output for each alert

Each alert must include:

* rule name;
* rule category;
* severity;
* affected entity;
* affected lines;
* calculation;
* source;
* explanation;
* confidence level;
* recommendation;
* status.

### Example alert

Alert type: Old outstanding commitment
Severity: Medium or High
Condition: commitment date older than six months and remaining amount greater than threshold
Explanation: “This commitment remains open more than six months after creation and still has a remaining amount of €X.”
Source: commitment file, line X
Recommendation: “Verify whether the service has received the goods/services, whether the invoice is pending, or whether the commitment should be closed.”

The rule must be deterministic and auditable. The AI may explain the alert, but the detection should initially come from explicit rules.

---

## 10. Conversational AI Module

The conversational AI module must allow users to ask business questions in natural language.

It must support:

* business-oriented chat;
* intent classification;
* retrieval of structured data;
* retrieval of documents through RAG;
* sourced answers;
* calculation explanations;
* refusal when data are unavailable;
* reformulation for different user profiles;
* generation of notes;
* generation of action plans;
* response export;
* conversation history.

### Required answer structure

A recommended answer should contain:

1. short answer;
2. key figures;
3. explanation;
4. sources;
5. limits;
6. cautious recommendation;
7. possible next actions.

### Mandatory guardrails

The assistant must:

* never provide a figure without a source;
* never invent a document;
* never invent missing data;
* explicitly state missing data;
* distinguish fact, analysis, and recommendation;
* include calculation details;
* cite source files or documents;
* include a confidence level;
* refuse out-of-scope requests;
* respect user permissions;
* log prompts and responses.

### Developer implication

The LLM should not directly query raw databases without control. A safer architecture is:

1. classify user intent;
2. identify required data;
3. check permissions;
4. retrieve structured data through controlled queries;
5. retrieve relevant documents through RAG if needed;
6. compute figures through deterministic functions;
7. pass summarized structured context to the LLM;
8. generate answer with strict instructions;
9. validate that the answer contains sources;
10. log the full interaction.

---

## 11. RAG and Document Management

The RAG module must allow the assistant to answer questions based on uploaded documents.

Documents to index may include:

* budgets;
* budget amendments;
* administrative accounts;
* financial statements;
* budget orientation reports;
* deliberations;
* public contracts;
* agreements;
* supporting documents;
* internal notes;
* procedures;
* M57 documentation;
* audit reports.

### RAG requirements

The RAG system must include:

* indexing by local authority;
* document metadata;
* access-right filtering;
* business-aware chunking;
* hybrid search;
* vector search;
* keyword search;
* citations;
* minimum relevance threshold;
* contradiction handling;
* source links;
* document versioning.

### Mandatory metadata

Each document should include:

* document id;
* local authority id;
* fiscal year;
* document type;
* title;
* upload date;
* uploader;
* source file;
* confidentiality level;
* service scope;
* version;
* indexed status.

### RAG answer behavior

If the retrieved evidence is weak, the assistant must say so. If documents contradict each other, it must identify the contradiction and avoid giving a definitive answer unless the hierarchy of documents is clear.

The assistant must never use documents from another local authority or another unauthorized service scope.

---

## 12. Dashboard Module

The dashboard module should provide clear visual summaries.

The MVP should include dashboards for:

* budget execution;
* commitments;
* mandates;
* suppliers;
* alerts;
* risks;
* service-level tracking.

Future dashboards may include:

* revenue titles;
* public contracts;
* debt;
* treasury;
* financial ratios;
* financial forecasting;
* elected-official summary.

### Budget dashboard

Should display:

* opened credits;
* committed amount;
* mandated amount;
* paid amount;
* available credits;
* execution rate;
* consumption rate by chapter;
* top drifting lines;
* under-consumed lines;
* risk alerts.

### Commitments dashboard

Should display:

* number of open commitments;
* total remaining committed amount;
* old commitments;
* commitments by service;
* commitments by supplier;
* commitments without contract;
* duplicated or suspicious commitments.

### Mandates dashboard

Should display:

* total mandates;
* rejected mandates;
* rejection reasons;
* rejected mandates by supplier;
* late mandates;
* high-value mandates.

### Supplier dashboard

Should display:

* top suppliers by amount;
* supplier concentration;
* duplicate suppliers;
* suppliers linked to rejected mandates;
* suppliers without contract reference if expected.

### Alert dashboard

Should display:

* alerts by severity;
* alerts by type;
* alerts by service;
* unresolved alerts;
* action status;
* export option.

---

## 13. Report and Export Generation

The product must generate operational deliverables, not only chat answers.

Expected exports:

* enriched Excel files;
* Word notes;
* PDF reports;
* anomaly reports;
* internal control reports;
* supplier sheets;
* contract sheets;
* service sheets;
* budget execution notes;
* action plans.

Each report must include:

* date;
* local authority;
* scope;
* fiscal year;
* source files;
* indicators;
* analysis;
* limits;
* recommendations;
* appendices.

### MVP report types

The first reports to implement should be:

1. budget execution note;
2. old commitments report;
3. supplier concentration report;
4. rejected mandates report;
5. internal control action plan;
6. executive summary for DGS;
7. simplified note for elected officials.

### Report generation flow

A recommended flow:

1. user asks question or selects report type;
2. system identifies required data;
3. system verifies data availability;
4. system computes indicators;
5. system retrieves sources;
6. system generates draft text;
7. system displays preview;
8. user can export to Word/PDF/Excel;
9. export is logged.

---

## 14. User Management, Roles, and Permissions

The system must support role-based access control.

Typical roles:

* super administrator;
* local authority administrator;
* finance director;
* chief administrative officer;
* elected official;
* finance agent;
* service manager;
* reader;
* external consultant;
* auditor.

### Permission requirements

Permissions must control:

* access to local authority data;
* access to service-level data;
* file upload;
* file deletion;
* import validation;
* dashboard access;
* chat access;
* report generation;
* user administration;
* rule configuration;
* audit log access.

### Tenant isolation

Tenant isolation is critical. Data from one local authority must never be visible to another.

Every database query must include tenant filtering. Every document retrieval query must include tenant filtering. Every vector search must be filtered by tenant and access rights.

This is a central architectural requirement.

---

## 15. Target Technical Architecture

The target architecture should be modular, API-first, secure by design, observable, and scalable.

### Suggested stack

Frontend:

* React;
* Next.js;
* TypeScript.

Backend:

* Python FastAPI or Node.js.

Data processing:

* Python;
* Pandas;
* Polars.

Database:

* PostgreSQL.

Vector database:

* Qdrant;
* Weaviate;
* pgvector.

File storage:

* S3-compatible storage;
* sovereign or European cloud storage.

AI:

* Mistral AI;
* embedding model;
* chat model.

Authentication:

* OAuth2;
* OpenID Connect;
* SSO;
* MFA.

Observability:

* logs;
* metrics;
* traces;
* cost monitoring;
* error tracking.

### Logical architecture

User
→ Frontend
→ Backend API
→ Authentication / RBAC
→ Business engine
→ PostgreSQL
→ Rules engine
→ AI orchestrator
→ Vector database / RAG
→ Mistral AI
→ Export module

Imports / APIs / connectors
→ Quality checks
→ Mapping
→ Raw storage
→ Normalized database

---

## 16. AI Architecture

The AI layer should be designed as an orchestrated system, not a direct LLM interface.

### Main components

1. intent classifier;
2. permission checker;
3. structured data retriever;
4. document retriever;
5. calculation engine;
6. prompt builder;
7. LLM response generator;
8. response validator;
9. citation checker;
10. logging module.

### AI options

The preferred model provider is Mistral AI or another European AI provider.

Possible deployment strategies:

* API-based for prototype and MVP;
* European or sovereign cloud for public-sector acceptability;
* open-weight model for stronger control;
* private hosting for sensitive deployments.

The architecture should avoid model lock-in. The LLM provider should be abstracted through a model interface.

### Anti-hallucination rules

The system must enforce:

* no figure without source;
* no invented source;
* no answer if data are unavailable;
* explicit limits;
* calculation details;
* clear confidence level;
* separate facts, analysis, recommendations;
* refusal outside scope;
* logs for audit;
* evaluation on test cases.

---

## 17. Security Requirements

Security must be integrated from the beginning.

Required security measures:

* TLS in transit;
* encryption at rest;
* secure file storage;
* MFA;
* SSO support;
* role-based access control;
* tenant isolation;
* audit logs;
* import logs;
* conversation logs;
* secret rotation;
* environment separation;
* backup strategy;
* disaster recovery plan;
* incident management;
* monitoring;
* reversibility;
* data retention policy.

### Sensitive data

The system may process:

* supplier data;
* employee-related financial data;
* debtor data;
* aid beneficiary data;
* contracts;
* invoices;
* supporting documents;
* internal financial notes;
* decision-making documents.

The design must therefore follow GDPR principles, including minimization, access control, traceability, and retention management.

### Main risks to mitigate

* data leakage;
* unauthorized access;
* AI hallucination;
* wrong interpretation of data;
* tenant confusion;
* poor source data quality;
* unexplainable answer;
* GDPR non-compliance;
* service unavailability.

---

## 18. Interoperability Strategy

The product must not depend on immediate API access from existing financial software vendors.

Integration should be progressive.

### Level 1 — Manual import

Users upload Excel or CSV files.

This is the MVP baseline.

### Level 2 — Automated file deposit

Files are periodically deposited in a secure space.

### Level 3 — Semi-automated connector

The system recognizes recurring formats from specific vendors.

### Level 4 — Partner API

Integration through vendor APIs when available.

### Level 5 — Native integration

The product becomes integrated with, or distributed alongside, existing business software.

### Priority data sources

The MVP should prioritize:

* budget;
* opened credits;
* commitments;
* mandates;
* suppliers;
* rejected mandates if available.

Later versions may include:

* revenue titles;
* public contracts;
* debt;
* treasury;
* supporting documents;
* remaining commitments;
* fixed assets;
* invoices.

---

## 19. UX/UI Requirements

The interface must be clear, modern, sober, professional, and suitable for the public sector.

It must serve both expert users and non-technical users.

### Main screens

The product should include:

1. home page;
2. secure login page;
3. main dashboard;
4. AI chat page;
5. data library;
6. file import page;
7. alerts page;
8. reports page;
9. history page;
10. administration page;
11. settings page;
12. user management page.

### Chat interface

The chat should include:

* question input;
* suggested questions;
* sourced answer;
* key figures;
* calculation details;
* source display;
* confidence level;
* action buttons;
* export button;
* reformulation button;
* note generation button;
* conversation history.

### UX principle

The product should display the synthesis first, then allow the user to drill down into details and sources.

For example:

1. show the main answer;
2. show key figures;
3. show explanation;
4. show source table;
5. show calculation formula;
6. allow export.

---

## 20. Import and Mapping UX

The file import workflow must be user-friendly because imported files may vary a lot.

Recommended workflow:

1. user uploads file;
2. system detects file type;
3. system previews rows;
4. system proposes data category;
5. system maps columns automatically;
6. user confirms or edits mapping;
7. system runs quality checks;
8. system displays warnings;
9. user validates import;
10. system stores raw and normalized data;
11. system generates import report.

The system should save mapping templates by local authority and by file type to reduce repeated manual work.

---

## 21. Logging and Auditability

The product must keep detailed logs.

Required logs:

* user login;
* file upload;
* import validation;
* column mapping;
* data transformation;
* rule execution;
* alert generation;
* chat question;
* AI response;
* sources used;
* export generation;
* permission changes;
* user administration changes.

Logs must be searchable and exportable for audit purposes.

For AI interactions, logs should include:

* user id;
* role;
* local authority id;
* question;
* intent;
* data retrieved;
* documents retrieved;
* model used;
* answer;
* citations;
* confidence level;
* timestamp.

---

## 22. Development Roadmap

### Phase 1 — Prototype

Objective: validate feasibility using Excel/CSV files.

Duration: 2 to 3 months.

Build:

* local import;
* basic mapping;
* structured model;
* first indicators;
* simple rules;
* experimental chat;
* basic sourced responses;
* initial dashboard;
* first export.

Deliverables:

* prototype;
* initial data dictionary;
* MVP backlog;
* first test dataset;
* UX mockups.

Validation:

* around 30 priority use cases tested;
* imported data correctly mapped;
* answers are sourced;
* basic alerts are explainable.

### Phase 2 — MVP Web Application

Objective: build a usable web app.

Build:

* authentication;
* file import workflow;
* mapping interface;
* PostgreSQL storage;
* dashboards;
* rules engine;
* AI chat;
* RAG;
* exports;
* user roles;
* import logs;
* conversation logs.

Validation:

* real or realistic anonymized data imported;
* dashboards stable;
* sourced answers generated;
* exports usable;
* tenant isolation tested.

### Phase 3 — Pilot Version

Objective: test with one to three local authorities.

Build:

* improved roles;
* service-level permissions;
* enriched rules;
* report templates;
* usage tracking;
* support tools;
* improved RAG;
* quality monitoring.

Validation:

* regular use by finance teams;
* issues logged;
* rule accuracy improved;
* user feedback integrated.

### Phase 4 — Industrialization

Objective: prepare for SaaS or private deployment.

Build:

* robust multi-tenancy;
* advanced monitoring;
* backup and recovery;
* scalable architecture;
* connector framework;
* onboarding workflow;
* advanced permissions;
* AI cost monitoring;
* security documentation;
* API documentation.

---

## 23. Developer Priorities

The most important development priorities are:

1. build a reliable ingestion pipeline;
2. preserve raw files and source traceability;
3. create a clean financial data model;
4. implement deterministic calculations;
5. build simple but explainable rules;
6. enforce tenant isolation everywhere;
7. design the AI as an orchestrated module;
8. require citations and sources;
9. log every import, alert, chat, and export;
10. generate usable reports.

The project should not start by building a sophisticated chatbot. It should start by structuring data and ensuring calculations are correct.

---

## 24. Recommended MVP Technical Backlog

### Epic 1 — Authentication and tenant setup

* Create local authority entity.
* Create users.
* Assign roles.
* Implement login.
* Implement tenant filtering.
* Implement service-level access where possible.

### Epic 2 — File import

* Upload Excel/CSV.
* Store raw file.
* Detect columns.
* Preview data.
* Map columns.
* Validate mapping.
* Normalize data.
* Save import batch.
* Generate import report.

### Epic 3 — Data model

* Create budget line model.
* Create commitment model.
* Create mandate model.
* Create supplier model.
* Create source reference model.
* Link all entities to tenant and import batch.

### Epic 4 — Business indicators

* Calculate execution rate.
* Calculate available credits.
* Calculate committed rate.
* Calculate mandated rate.
* Calculate remaining commitment.
* Calculate supplier concentration.
* Calculate old commitments.
* Calculate rejected mandate statistics.

### Epic 5 — Rules engine

* Create rule registry.
* Implement old commitment rule.
* Implement insufficient credit rule.
* Implement abnormal consumption rule.
* Implement supplier concentration rule.
* Implement duplicate supplier rule.
* Implement rejected mandate rule.
* Generate alerts.
* Store alerts.
* Display alerts.

### Epic 6 — Dashboards

* Budget dashboard.
* Commitments dashboard.
* Mandates dashboard.
* Suppliers dashboard.
* Alerts dashboard.
* Filters by fiscal year, service, chapter, supplier.

### Epic 7 — AI chat

* Intent classification.
* Controlled data retrieval.
* Calculation explanation.
* Source citation.
* Answer generation.
* Refusal behavior.
* Conversation history.
* Export response.

### Epic 8 — RAG

* Upload documents.
* Extract text.
* Chunk documents.
* Generate embeddings.
* Store vectors.
* Retrieve documents by tenant and permissions.
* Display citations.
* Include sources in AI answers.

### Epic 9 — Reports

* Budget execution note.
* Old commitments report.
* Supplier concentration report.
* Rejected mandates report.
* Internal control action plan.
* Export to Word/PDF/Excel.

### Epic 10 — Audit and monitoring

* Import logs.
* User logs.
* AI logs.
* Rule execution logs.
* Export logs.
* Error logs.
* Basic admin dashboard.

---

## 25. Key Technical Principles

The product must follow several non-negotiable principles.

### Source-first architecture

Every number must be linked to a source file, source row, source document, or computed formula.

### Deterministic calculations

Important financial calculations should be performed by backend services, not invented by the LLM.

### AI as explanation layer

The LLM should explain, summarize, reformulate, and draft notes. It should not be the only place where business logic lives.

### Strict tenant isolation

Every request, query, vector search, document access, and export must be filtered by local authority and user permissions.

### No silent failure

If data are missing, the system must say so. If an import is partial, the system must show warnings. If the AI lacks evidence, it must refuse or qualify the answer.

### Auditability

The product must be explainable after the fact: who uploaded what, which data were used, which rule generated which alert, which source supported which answer.

---

## 26. Final Developer Summary

The product to build is a secure vertical financial intelligence platform for local governments.

The first version must allow users to upload Excel/CSV files, map them to a normalized financial data model, compute budget and accounting indicators, detect simple anomalies, query data through an AI assistant, display dashboards, and generate sourced reports.

The most important modules are:

* ingestion;
* mapping;
* financial data model;
* rules engine;
* AI orchestration;
* RAG;
* dashboards;
* exports;
* user roles;
* tenant isolation;
* logging.

The developer should prioritize correctness, traceability, and security over advanced AI sophistication.

The MVP should prove that the system can answer practical questions such as:

* What is the budget execution rate?
* Which credits may be insufficient?
* Which commitments are old and still open?
* Which suppliers concentrate spending or anomalies?
* Which mandates were rejected?
* Can the system generate a sourced note for management or elected officials?

The long-term product can later add connectors, APIs, forecasting, advanced scoring, debt, treasury, procurement, and broader public finance analysis.

But the first technical objective is clear: build a reliable, secure, source-based assistant that transforms imported local government finance data into dashboards, alerts, explanations, and reports.

হাই Kauser,

আমি 1 মিনিট ট্রেডিংয়ের জন্য একটি high-accuracy Quotex trading signal/bot system বানাতে চাই।

প্রয়োজনীয়তা:
- প্রতিটি ক্যান্ডেল বিশ্লেষণ
- Price Action + Trend filtering
- Support & Resistance (SNR) ভিত্তিক এন্ট্রি
- OTC এবং normal market support
- দ্রুত Buy/Sell সিগনাল
- কম repainting এবং high-probability setup
- Risk management system
- Telegram alerts (optional)

আপনি কি এ ধরনের সিস্টেম বানাতে পারবেন?

দয়া করে জানান:
1. আনুমানিক খরচ কত হবে?
2. ডেলিভারি টাইম কত লাগবে?
3. কোনো ডেমো বা আগের কাজ আছে কি?

ধন্যবাদ

হাই Kauser,

আমি 1 মিনিট ট্রেডিংয়ের জন্য একটি high-accuracy Quotex trading signal/bot system বানাতে চাই।

প্রয়োজনীয়তা:
- প্রতিটি ক্যান্ডেল বিশ্লেষণ
- Price Action + Trend filtering
- Support & Resistance (SNR) ভিত্তিক এন্ট্রি
- OTC এবং normal market support
- দ্রুত Buy/Sell সিগনাল
- কম repainting এবং high-probability setup
- Risk management system
- Telegram alerts (optional)

আপনি কি এ ধরনের সিস্টেম বানাতে পারবেন?

দয়া করে জানান:
1. আনুমানিক খরচ কত হবে?
2. ডেলিভারি টাইম কত লাগবে?
3. কোনো ডেমো বা আগের কাজ আছে কি?

ধন্যবাদ

হাই Kauser,

আমি 1 মিনিট ট্রেডিংয়ের জন্য একটি high-accuracy Quotex trading signal/bot system বানাতে চাই।

প্রয়োজনীয়তা:
- প্রতিটি ক্যান্ডেল বিশ্লেষণ
- Price Action + Trend filtering
- Support & Resistance (SNR) ভিত্তিক এন্ট্রি
- OTC এবং normal market support
- দ্রুত Buy/Sell সিগনাল
- কম repainting এবং high-probability setup
- Risk management system
- Telegram alerts (optional)

আপনি কি এ ধরনের সিস্টেম বানাতে পারবেন?

দয়া করে জানান:
1. আনুমানিক খরচ কত হবে?
2. ডেলিভারি টাইম কত লাগবে?
3. কোনো ডেমো বা আগের কাজ আছে কি?

ধন্যবাদ

This posting is open to emerging talent and seasoned developers alike. Years of experience do not matter — if you are a self-taught developer, a recent graduate, or a junior who has quietly shipped impressive production work, we want to hear from you. The specification is complete, the architecture is decided, and a deployable starting point is provided, which means the right person can deliver this project at a competitive price without compromising on quality.
We actively encourage the use of AI coding tools — Cursor, GitHub Copilot, Claude, or any assistant that helps you move faster and smarter. We expect you to use them. What we do not accept is unreviewed, untested output: every line that goes into production must be understood, validated and owned by you. AI speeds up the work; your expertise ensures it is correct.
What matters is your portfolio, not your CV. If the stack below matches your background, your work proves it, and you know how to combine your own skills with the tools available to you today — apply.

Project type
Full-stack subscription platform — web application
Engagement
Fixed price · milestone-based · 11 milestones

Duration
Estimated 6–8 weeks
Team size1 senior developer or small team (max. 2)
Location
Remote · EU timezone preferred (CET ±2h)

Confidentiality
NDA required before full specification is shared

1. What We Are Building
A modern, subscription-based creator-supporter platform in the sports and community sector. Creators publish exclusive content and sell personalised items directly to their fan community. Supporters subscribe monthly and access a protected feed. An admin backend handles accreditation, content moderation and payout management.
The platform handles real money, real subscriptions and legally compliant invoicing. It must be secure, GDPR-compliant, hosted entirely in the EU, and built to last.

Why this project is worth your time: Everything is already designed and specified. At project start you receive: full architecture (C4 diagrams), complete database schema, 11 high-fidelity UX screens, a clickable prototype, a deployable Stripe webhook handler (400+ lines TypeScript), and concrete acceptance criteria per milestone. Your job is clean implementation.


2. What We Expect
This is not an entry-level project. We need a developer who has shipped production applications with the specific stack below. If you have not worked with Stripe Connect Express, Supabase RLS and Next.js App Router in a real production context, this posting is not for you.

Immediate disqualifiers: No Stripe Connect Express experience · No Supabase Row Level Security experience · No production Next.js App Router project · TypeScript with any types or disabled strict mode · CMS-first approach (WordPress, Drupal, Strapi, etc.)


3. Required Skills
Frontend

Next.js 15 — App Router, React Server Components, SSR/ISR — production experience required
TypeScript strict mode throughout — no any, no ts-ignore
Tailwind CSS + shadcn/ui — component-based UI, responsive across all breakpoints
Clear understanding of client vs. server component boundaries

Backend & Database

Supabase PostgreSQL with Row Level Security — you must have written RLS policies in production
Prisma ORM — schema design, relations, migrations, type-safe queries
Next.js API Routes with Zod input validation on every endpoint
Supabase Auth — JWT sessions, email double opt-in, route-level middleware
Rate limiting on authentication endpoints

Stripe — Hard Requirement

Stripe Checkout Sessions — creation, metadata, redirect handling
Stripe Connect Express — marketplace model, creator onboarding, direct payouts
Stripe Webhooks — signature verification, event handling, idempotency
Stripe Subscriptions — lifecycle, cancellation, status sync, past-due handling
Note: No demonstrable Stripe Connect experience = application not considered.

Media & Storage

Cloudflare R2 — object storage, signed URLs (time-limited access), CORS configuration
Cloudflare Stream — HLS video delivery, no direct download exposure
tus protocol + Uppy client — chunked, resumable uploads up to 2 GB
Content protection — preventing indexing, scraping and unauthorised media access

Email & Documents

Resend or equivalent transactional email service with React Email templates
Batch email delivery to subscriber lists
PDF invoice generation with @react-pdf/renderer — sequential numbering, VAT-compliant

Infrastructure & Compliance

Vercel — EU region deployment, environment config, CI/CD via GitHub
GDPR compliance — EU-only hosting, soft-delete, data export and deletion endpoints
Security headers — X-Robots-Tag, Content-Security-Policy, HSTS
robots.txt — including AI crawler blocking (GPTBot, ClaudeBot, CCBot, etc.)
Cloudflare — Bot Management, Turnstile, AI Bot Protection
Sentry EU + PostHog EU — error monitoring and privacy-compliant analytics


4. Preferred Experience
ExperienceWhy it mattersMarketplace or multi-tenant platformsData isolation per creator, role-based access, per-creator payout logicEU eCommerce complianceGoBD-compliant invoice numbering, VAT handling for digital servicesRealtime features with SupabaseLive feed updates via Postgres channels, WebSocket subscription managementLarge file upload pipelinestus server-side handling, progress tracking, resumable sessionsStripe revenue share and fee splitsAutomatic platform fee deduction on Connect transactionsContent moderation toolingAdmin workflows for removal, audit trails, creator notificationsSports, creator or fan platformsFamiliarity with the user mental model and content patterns

5. Project Scope — Milestone Overview
The project is divided into 11 milestones. Each is independently reviewable and compensated upon written acceptance. Full task lists and acceptance criteria are shared after NDA.

(The milestone plan and compensation structure outlined above serve as a working baseline. Both will be reviewed and refined together with the selected candidate before project start — we are open to adjustments that reflect your working style, availability and preferred payment rhythm, as long as the overall structure of independent, verifiable deliverables is maintained.)

Milestone
What it coversWeight
M-01 SetupNext.js, Supabase, Vercel EU, Cloudflare R2/Stream, Stripe, Resend, Sentry, PostHog5 %
M-02 Database & RLS13 tables, Prisma schema, all migrations, 15+ RLS policies, DB triggers, seed script9 %
M-03 Auth & Roles3-role system, middleware route guards, email double opt-in, rate limiting8 %
M-04 Creator Onboarding5-step accreditation flow, document upload, admin review queue, Stripe Connect Express10 %
M-05 Landing PageSSR creator grid, profile pages, blurred feed preview, SEO, security headers7 %
M-06 Stripe & WebhooksCheckout sessions, 7 webhook event handlers, idempotency, subscription management11 %
M-07 Content & Feedtus/Uppy upload, consent flow, HLS delivery, RLS-filtered feed, Realtime15 %
M-08 Shop & PWYWItem creation, pay-what-you-want pricing, server-side validation, signed URL delivery11 %
M-09 Admin DashboardAccreditation queue, user management, content moderation, suspensions, payouts10 %
M-10 Email & Invoices16 email templates, PDF invoices, sequential numbering, notification preferences7 %
M-11 GDPR & Launchrobots.txt, security headers, Cloudflare bot protection, GDPR data endpoints7 %

Total working days27–29 days · approx. 6–8 weeks at 5h/dayLargest milestoneM-07 Content & Feed — 15 % of total, 4 daysCritical pathM-01 → M-02 → M-03 → M-06 → M-07Payment triggerWritten acceptance per milestone · 5 business day review windowAdvance paymentNone — first payment after M-01 acceptance only

6. What You Receive at Kickoff
DeliverableWhat it containsRequirements document60+ pages, complete functional and non-functional requirements, acceptance criteria per milestoneArchitecture packageC4 diagrams Level 1–3, container map, component breakdownDatabase schemaFull Prisma schema, all 13 tables, RLS policies, enum types — ready to migrateUX package11 high-fidelity screens + clickable HTML prototype covering every user flowStripe webhook handlerDeployable TypeScript file, 400+ lines — all 7 events, idempotency, audit loggingTarget environmentEU servers configured, all accounts set up — no infrastructure decisions neededAcceptance criteriaPer-milestone test checklist — binary pass/fail, no interpretation required

Zero dependency after project end: Full Git repository with history handed over on final acceptance. All credentials and accounts owned by the client. A 2-hour handover session ensures any future developer can pick up where you left off.


7. How to Apply
Required in your proposal

Link to a live Next.js App Router project or public GitHub repository
One concrete Stripe Connect Express example — what you built, what the marketplace model was, any complexity you handled
Brief description of a Supabase RLS policy you wrote yourself — what table, what rule, how you tested it
One reference for a GDPR-compliant, EU-hosted project
Fixed-price proposal — either a single total or per-milestone breakdown
Availability in hours per week and earliest start date

Optional — strengthens your application

Cloudflare Stream or R2 experience — describe a project or share a snippet
Example React Email template or PDF generated with @react-pdf/renderer
Experience with subscription billing, recurring payments or marketplace fee models

Selection process

We respond to all shortlisted applications within 5 business days
Shortlisted candidates receive a brief technical questionnaire (20–30 minutes)
Selected candidate signs NDA and receives the full specification package immediately
Project starts within 2 week of NDA signing

We are seeking a skilled developer to create a tournament website that will eventually be scaled into a mobile and desktop app. The ideal candidate will have experience in both web and app development, ensuring a seamless transition between platforms. The project requires a strong understanding of user interface and user experience to ensure the app is engaging and easy to use. If you have a passion for developing dynamic and interactive applications, we would love to hear from you!

• Describe your recent experience with similar projects
• Include a link to your GitHub profile and/or website

I built a voice AI customer support agent for my retail business using Make.com and Vapi. It was working fine for about 3 weeks then suddenly stopped. When customers call in, the call connects but the agent either hangs silently, gives a generic response, or drops the call entirely. I haven't changed anything major maybe a few small field updates. I need someone who knows both Make.com scenario logic AND Vapi call flows to get in, diagnose it, and fix it without breaking what was already working. This is urgent it's customer-facing. Budget $10–15 fixed. If you know what you're doing this should not take long.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
PROJECT OVERVIEW
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

We are building a multi-tenant, AI-native operating system for the NDIS (National Disability Insurance Scheme) and Aged Care sector in Australia.

This is a large, long-term build. We are not looking for someone to build a website, a landing page, or a simple CRUD application. We are building an enterprise-grade SaaS platform from the ground up that will eventually serve thousands of organisations across the Australian healthcare sector.

The platform consolidates dozens of separate tools currently in use across multiple registered healthcare organisations into 18 fully integrated modules on a single architecture — with a purpose-built AI agent layer that automates high-volume operational and compliance tasks while keeping humans in control of every consequential decision.

We are looking for someone who wants to build something significant — not fill a short contract. The right person will grow with this platform for years. For the right candidate, this engagement has no defined end date. There is a genuine long-term association on offer here, with the platform growing in scope, complexity, and commercial scale over time.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
WHAT WE ARE BUILDING
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

An18-module platform covering:

• Staff Management & HR — full employee lifecycle, compliance tracking, credential management, Award calculations
• Recruitment Pipeline — AI-powered 6-stage hiring pipeline with agent-driven application screening, scoring, interview scheduling, and onboarding
• Participant Management — NDIS participant records, support plans, funding tracking, progress notes
• Compliance & Screening — real-time compliance monitoring across NDIS Practice Standards quality indicators
• Incidents & Quality — incident reporting, complaint management, risk register
• Form Creator & Workflow Engine — meta-configurable form builder with conditional logic, e-signature, and automation
• Rostering & Scheduling — shift management, availability, calendar integration
• Properties, Fleet & Assets — facility and vehicle management
• Purchasing & Finance — purchase orders, accounting integration, payment processing
• Reports & Analytics — real-time operational and compliance dashboards
• Mobile Application — React Native field worker app with offline mode
• Microsoft 365 Integration — Outlook, Calendar, SharePoint, Teams
• SaaS Engine — multi-tenant billing, self-service onboarding, subdomain system
• And lots more

The AI agent layer runs across all modules. It is not a chatbot. It is a structured system of background workers that call the LLM API to perform specific, auditable tasks — document generation, application scoring, compliance monitoring, employment agreement population, scheduling — with configurable confidence thresholds and mandatory human approval at every consequential decision point.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
INTEGRATIONS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

PRISM is integration-heavy by design. The platform must connect with a wide range of external systems. Confirmed integrations include — but are not limited to:

AI & Automation
• Anthropic Claude API — AI agent layer, document generation, scoring, analysis
• OpenAI API — fallback and specialist AI tasks
• Google Gemini API — form generation and analysis
• WhatsApp Business API — automated client and participant communications
• Twilio — SMS notifications, verification, two-factor authentication

Microsoft Ecosystem
• Microsoft Graph API — Outlook email sync, Calendar, SharePoint, Teams notifications
• Microsoft 365 — embedded Word and Excel editing, document management
• Azure Active Directory — SSO and identity federation for enterprise clients

Finance & Payments
• Xero API — accounting, invoicing, payroll sync
• Airwallex — payment processing and multi-currency
• Stripe — SaaS subscription billing, plan management, invoicing
• MYOB API — accounting integration (alternate to Xero)

Government & Compliance
• NDIS Myplace Provider Portal — plan and funding data (where API access is available)
• Medicare / Services Australia — bulk billing and claim processing
• ATO Single Touch Payroll (STP) — payroll reporting compliance
• Workers Compensation APIs — state-based reporting

Recruiting & HR
• Idibu API — multi-board job advertisement publishing (Indeed, LinkedIn, Reed, Totaljobs, Seek)
• Seek API — direct job board integration
• LinkedIn API — talent sourcing and job posting
• DocuSign / PRISM Signing Module — e-signature for employment agreements and consents
• VEVO — visa and right to work verification
• ACIC — national police check integration

Communications & Productivity
• SendGrid — transactional email, bulk notifications, template management
• • Zoom API — video consultation scheduling and links
• Google Calendar API — alternative to Microsoft 365 calendar
• Slack API — internal team notifications

Mapping & Location
• Google Maps API — participant and staff address verification, route planning
• HERE Maps API — alternative routing and proximity matching

Storage & Files
• Cloudflare R2 — S3-compatible file storage, no egress fees
• AWS S3 — document and media storage
• SharePoint — enterprise document management via Microsoft Graph

Analytics & Monitoring
• Google Analytics — product usage analytics
• Mixpanel — user behaviour tracking
• Sentry — error tracking and alerting
• AWS CloudWatch — infrastructure monitoring
• Datadog — application performance monitoring (APM)

Security & Identity
• Auth0 — identity management and SSO for SaaS tier
• Google reCAPTCHA — bot protection
• HaveIBeenPwned API — credential breach checking

This list grows as the platform scales. The Tech Lead must be comfortable designing a robust, versioned, fault-tolerant integration architecture that can absorb new integrations without destabilising the core platform.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
TECHNOLOGY STACK —
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Backend: Node.js — REST API, BullMQ job queue, webhook engine
Frontend: React SPA with Tailwind CSS
Mobile: React Native (iOS + Android)
Database: PostgreSQL on AWS RDS with read replica
Cache / Queue: Redis on AWS ElastiCache — BullMQ for all agent tasks
File Storage: Cloudflare R2 (S3-compatible)
Auth: JWT with 2FA, role-based access control to field level
AI: Anthropic Claude API
CI/CD: GitHub + GitHub Actions
Hosting: AWS Sydney

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
YOUR ROLE
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

This is a Tech Lead role. You will own:

• All architectural and engineering decisions
• Database schema design and evolution
• REST API design and implementation
• BullMQ agent worker implementation — Claude API calls, confidence scoring, hard-block enforcement, audit logging
• Integration architecture — designing a fault-tolerant, versioned layer that connects all external systems listed above
• AWS infrastructure — EC2 / RDS / ElastiCache / S3 / CloudFront setup and management
• CI/CD pipeline — GitHub Actions
• Security and data compliance — Australian Privacy Act, NDIS Practice Standards, data sovereignty
• Recruiting, onboarding, and managing the junior offshore development team
• Code quality standards, code review, and engineering culture
• Technical documentation — architecture docs, API docs, developer onboarding

You will report directly to the Managing Director. You will have full technical autonomy within the agreed architecture.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
LONG-TERM ASSOCIATION
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

This is not a project with a defined end date. The platform has six build phases and a commercial pathway into the broader NDIS and Aged Care SaaS market. Once internal deployment is complete, the platform will be offered to thousands of providers across Australia.

For the right person, this means:

• Growing from Tech Lead into the senior-most technical figure on an expanding platform
• Leading a growing team as the organisation scales
• Being part of the commercial expansion as the platform moves into SaaS
• Building a body of work that has real sector-wide impact in Australian healthcare

We are not looking for someone to build it and hand it over. We are looking for someone who wants to own it.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
WHAT WE NEED IN YOU — NON-NEGOTIABLE
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Do not apply if you cannot specifically address all of the following:

1. You have built and shipped a multi-tenant SaaS platform in production — not a prototype, not a side project. A real product with real users and real data.

2. You have deep, hands-on PostgreSQL experience — schema design, indexing, query optimisation, migrations. Not just "I have used a database."

3. You have built REST APIs in Node.js at production scale — authentication, rate limiting, versioning, error handling.

4. You have worked with AWS in a production environment — EC2, RDS, ElastiCache, S3, IAM. Not just Vercel or Firebase.

5. You have experience managing or leading a development team — code reviews, task delegation, technical mentoring.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
STRONGLY PREFERRED
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

• Experience implementing agentic AI workflows — LLM API calls, background job queues, structured JSON outputs, confidence scoring
• Experience with BullMQ or a comparable job queue in production
• Microsoft Graph API experience — Outlook, Calendar, SharePoint
• Experience designing integration architecture across many third-party APIs
• Experience in healthcare, compliance, or regulated data environments
• Experience hiring and managing offshore developers

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
HOW TO APPLY — READ THIS CAREFULLY
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

We will not read generic proposals. To be considered, your proposal must answer both of the following questions specifically. Proposals that do not address both will not receive a response.

QUESTION 1:
Describe a multi-tenant SaaS application you have built. How did you implement tenant data isolation at the database level? What was your schema approach and why did you choose it?

QUESTION 2:
How would you architect a background job queue for AI agent tasks where some tasks auto-proceed and others pause for human review based on a confidence threshold? What technology would you use and how would you handle failures and retries?

If you can answer both questions clearly and specifically, we want to hear from you.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
PROJECT DETAILS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Duration: Long-term — no defined end date for the right person
Engagement: Part-time to start (minimum 3 days/week), growing to full-time equivalent as the build accelerates
Budget: Discussed based on experience — we are serious about this project and will pay for the right person
Sector: NDIS and Aged Care — Australian healthcare

This is a large project. If you are looking for a short contract, this is not it. If you want to build something that matters and grow with a platform that will serve thousands of care workers and the vulnerable Australians they support — let's talk.

• Describe a multi-tenant SaaS application you have built. How did you implement tenant data isolation at the database level? What was your schema approach and why did you choose it?
• How would you architect a background job queue for AI agent tasks where some tasks auto-proceed and others pause for human review based on a confidence threshold? What technology would you use and how would you handle failures and retries?

We are looking for a technical AI automation specialist to help us build a mostly automated AI video generation system for short-form content production.

This is not a traditional video editing role. We need someone who can build the backend workflow, connect tools/APIs, organize cloud storage, create reusable prompt systems, and help us automate the production of AI-generated video clips.

The goal is to generate daily batches of 5–10 second AI video clips for our Noomizo Skully content system. These are stylized, character-based AI videos. The clips do not need to be Hollywood-level cinematic, but they must look polished, consistent, and usable for social media.

Our biggest priorities are:
Strong character consistency across clips
Near-full automation
Repeatable daily clip production
Clean cloud folder organization
Easy handoff to our human video editor
Our video editor will manually stitch the clips together into final videos. Your job is to build the system that generates, organizes, watermarks, and stores the clips.

Workflow We Want to Build:
The system should be able to:
Generate short-form video ideas.
Turn those ideas into short scripts.
Break each script into shot lists.
Convert each shot into seed image prompts.
Generate seed images while keeping the character consistent.
Convert seed images into image-to-video prompts.

Generate multiple 5–10 second clip variations, such as:
Hook clips
Body clips
Ending clips
CTA clips
Automatically add our brand @name as a watermark to each clip.
Upload the clips to cloud storage.

Organize clips into clear folders, such as:
Hooks
Body Clips
Endings
CTAs
Approved Clips
Rejected Clips
Final Exports

What We Need From You:
We need someone who can help us set this up from start to finish, including:
Recommending the best tool stack
Building the automation workflow
Connecting AI APIs
Setting up cloud storage
Creating folder structures and naming conventions
Creating reusable prompt templates
Setting up watermarking
Building a basic approval/rejection system
Documenting the process so our team can use it without technical help

Preferred Tools / Experience:
Experience with some of the following is preferred:
n8n, Make, Zapier, or similar automation tools
ChatGPT / OpenAI API
Claude
fal.ai
Seedance
Runway
Kling
AI image generation workflows
AI image-to-video workflows
Google Drive, Dropbox, Supabase, AWS, or similar cloud storage
Airtable, Notion, Google Sheets, or similar tracking systems
FFmpeg or other tools for watermarking/video processing
API integrations
Prompt engineering
Character consistency workflows

Deliverables:
By the end of the project, we want:
A working automated AI clip generation workflow
A cloud folder system that organizes generated clips
Reusable prompt templates for ideas, scripts, shot lists, seed images, and video prompts
Automated watermarking with our @name
Clip categorization by type: hook, body, ending, CTA, etc.
A simple approval/rejection workflow
Documentation/SOPs showing how to run and maintain the system
Recommendations for scaling the system later

Ideal Candidate:
The ideal freelancer is technical, automation-minded, and familiar with AI content generation pipelines. You do not need to be a professional video editor, but you should understand how AI video tools work and how to create a repeatable workflow around them.
We are looking for someone who can think through the full system, not just connect one or two tools. The final result should reduce manual work and give our editor a reliable library of organized AI-generated clips to work from every day.

We are seeking a skilled and forward-thinking AI Developer to build a custom AI system for an innovative project. This role involves designing, developing, and implementing intelligent solutions tailored to specific business needs. If you are passionate about artificial intelligence and enjoy building impactful systems from scratch, we would love to hear from you.

Key Responsibilities:
Design and develop a custom AI-powered system
Build, train, and optimize machine learning models
Integrate AI solutions with existing platforms and APIs
Work closely with stakeholders to understand project requirements
Ensure scalability, performance, and security of the system
Provide documentation and technical support

Project Goal
A module utilizing AI/LLM technology is required for the automatic analysis of incoming email messages (reports) related to new construction processes / construction investments.
The system should:
1. analyze the email content (email language - Polish)
2. extract key information,
3. assess the lead priority,
4. assign the lead to the appropriate regional sales representative based on the investment location,
5. verify whether an analysis related to the same investment has already been performed within the last 3 months (elimination of duplicate information coming from different sources),
6. return structured data in JSON format for further processing by another system.
________________________________________
Functional Scope
1. Email Retrieval
Source:
• IMAP / Microsoft 365 / Gmail API (to be determined)
The module should:
• retrieve new messages,
• analyze the email subject and body,
• analyze PDF/DOCX attachments.
________________________________________
2. AI-Based Content Analysis
The system should automatically detect and structure, among others:
• type of investment,
• investment stage,
• construction site location/address,
• investor or company name,
• project scale,
• keywords defining priority,
• contact details of persons and companies responsible for the investment.
The system should extract and structure contact data found in the email body or attachments, including:
• company name,
• contact person’s full name,
• job title/role,
• phone number,
• email address,
• company/investment address (if available).
The data should be returned in a structured JSON format.
________________________________________
3. Priority Assignment
The system should assign a scoring / priority level:
• HIGH
• MEDIUM
• LOW
Based on:
• investment size,
• type of investment,
• keywords,
• project stage,
• presence of specific contact details,
• location,
• other configurable rules.
The rules should be editable in the future.
________________________________________
4. Geographic Routing
Based on the investment address or location, the system should assign the lead to the appropriate regional/area representative.
Example:
• Wielkopolskie and Śląskie regions → Region A
• Dolnośląskie region → Region B
• Mazowieckie region → Region C
The address may be provided as:
• full address,
• city name,
• postal code,
• voivodeship/region name.
The system should support ambiguous or incomplete location data.
________________________________________
5. Output – JSON
The system should return data in a structured JSON format.
The final JSON structure/format will be agreed during the implementation phase.
________________________________________
6. Expected Result
A production-ready module that:
• receives emails,
• automatically analyzes them,
• classifies them,
• assigns the appropriate region,
• returns JSON/API output with structured information.
________________________________________
Additional Information
Sample emails and reports will be provided.

We've got a customer support bot build in Intercom for our E-comm store. We've been continuously adding features to it over the last ~12 months or so.

It knows hot answer questions about our products, it knows how to recommend other products, etc.

We also created data connector that lets the AI bot answer order/tracking related questions. We run a couple Shopify stores (different stores for different countries) and The bot automatically recognizes the person if they reach out to us on email (links their email to Shopify account). In that case, the lead connector works fine.

The problem happens when a non logged-in user uses our chatbot on the website. We don't know their email and order ID, therefore Shopify won't recognize them and the data connector will fail. We're looking to automate email + order ID collection for users, where shopify data isn't available for them. We then want to continue towards passing the data to data connector if needed, once that's done.

Ideally, I'd like someone to hop on a call with me and help me resolve this live in our sessions. Please apply only if you feel like you can resolve this without learning everything for 5 hrs first.

Thanks!

# AI Automation Consultant – Claude.ai Agent Development

**Job Title:**
AI Automation Consultant / Claude.ai Agent Developer

**Job Type:**
Independent Contractor / Freelance

**Project Overview:**
We are seeking an experienced AI automation consultant who specializes in building intelligent workflow agents using [Claude.ai](https://claude.ai?utm_source=chatgpt.com) and related AI tools. Our company manages high volumes of pricing, sourcing, and operational tasks across multiple vendor and backend systems, and we are looking to streamline these workflows using AI-powered agents and automation.

The ideal candidate should understand how to build practical AI agents that can interact with multiple websites, gather information, compare data, and assist our internal team with repetitive operational tasks.

## What We Need

Currently, our staff manually logs into multiple vendor and supplier websites to:

* Check product availability
* Compare pricing
* Verify stock levels
* Gather order information
* Assist with quoting and procurement workflows

In many cases, our team must access 5 or more different platforms for a single order. We want to automate and streamline these processes using AI agents and backend automation systems.

## Scope of Work

We are looking for someone who can:

* Build AI agents using Claude.ai and/or similar AI frameworks
* Create automations that interact with multiple websites and backend systems
* Help streamline procurement and quoting workflows
* Build tools that can:

* Log into vendor portals
* Pull inventory and pricing data
* Compare supplier information
* Present consolidated results to our team
* Recommend the best architecture and AI stack for long-term scalability
* Help integrate APIs, browser automation, and AI workflows where needed
* Potentially integrate with internal systems such as spreadsheets, portals, or databases

## Preferred Skills

* Experience with Claude.ai, AI agents, and workflow automation
* Strong understanding of browser automation tools
* Experience with:

* Python
* APIs
* Web scraping/automation
* AI orchestration tools
* Backend integrations
* Familiarity with tools such as:

* Playwright
* Selenium
* LangChain
* MCP servers
* Zapier
* Make.com
* OpenAI / Anthropic APIs
* Experience building AI-powered business process automations
* Strong communication and consulting skills

## Ideal Candidate

We are looking for someone who is:

* Practical and solutions-oriented
* Experienced in operational workflow automation
* Able to recommend scalable AI solutions
* Comfortable working with multiple systems and vendor platforms
* Able to think like both an engineer and business consultant

## To Apply

Please provide:

* Examples of AI agents or automation systems you have built
* Relevant experience with Claude.ai or AI workflow automation
* Your recommended approach for this type of project
* Estimated timeline and hourly/project rates
* Any case studies or portfolio examples

We are seeking a specialist to automate our business processes, focusing on answering questionnaires, reducing paperwork, and creating summarized documents. The ideal candidate will have experience in business management and analysis, with a strong ability to streamline operations and improve efficiency. This role requires a proactive approach to identifying areas for automation and implementing effective solutions. PLEASE PROVIDE SAMPLES OF WORK IN YOUR PROPOSAL

We’re seeking a highly experienced Microsoft Copilot Developer to build a robust, production-ready solution for a European client. This project will involve creating a voice-driven Copilot agent within Microsoft Teams that can pull content from an on-premise file share and curated manufacturer websites, interact with users iteratively, and populate a Word document template before saving it back to the correct job folder on the file share.

This is a hands-on, delivery-focused role. You will be collaborating closely with the engagement lead throughout the development process, with regular check-ins and joint testing sessions to ensure seamless delivery.

Required Skills

The following skills are essential, and candidates must have hands-on experience in these areas. Without this, we will not be able to consider your application:

Copilot Studio: Expertise in building declarative agents, handling Entities, dynamic URL variables, publishing to Teams, and designing system prompts.
Microsoft Graph Connector Agent: Experience configuring agents to interact with on-premise file shares.
On-Premises Data Gateway: Familiarity with configuration and integration with Power Automate.
Power Automate: Proficiency in creating cloud flows within Dataverse Solutions, utilizing the Word Online (Business) connector with content controls, and handling on-premise file writes.
Microsoft 365 Administration: Experience with DLP policy reconciliation and tenant-level configuration.
Proven Copilot deployment in live environments: Deployment evidence is required—years of experience alone are insufficient.

Note: A combination of Copilot Studio, Graph Connector, and On-Premises Data Gateway expertise is highly specialized. General Power Platform experience will not be sufficient to meet the needs of this project.

Desirable Skills
Power Apps development experience
Advanced knowledge of Dataverse (beyond Solutions)
SharePoint architecture experience
Data structuring for Copilot readiness and data cleansing
Comfortable providing technical assessments during pre-sales calls

Setting up AI agents in Copilot and creating power automation, crm workflow integrations dataverse
My husband is a builder on the GOld Coast and he has great difficulty dealing with all administrative task:
1. emails (Clearing /sorting /responding)
2. estimating (working out prices for jobs / quoting jobs /working out variations )
3. Project management : scheduling and ordering materials

Python ML Developer for Financial Data Analysis Tool - Hourly $10-15/hr

Experienced ML engineer needed to build a Python-based tool for analyzing financial market data and generating trading signals.

Project Scope
Develop ML models (LSTM, Random Forest, etc.) for time-series prediction

Integrate market data APIs (Alpaca, Polygon, or similar)

Implement backtesting framework with performance metrics

Add basic risk management features (position sizing, stop-loss logic)

Create clean, documented codebase with testing

Required Skills
Python (Pandas, NumPy, Scikit-learn, TensorFlow/PyTorch)

Financial data handling and technical indicators

API integration and real-time data processing

Backtesting and quantitative analysis experience

To Apply
Please share:

Relevant portfolio/projects (GitHub links)

Experience with financial/ML tools

Your proposed approach and timeline

Long-term opportunity for strong performers. Interviews starting today

Hello!

Im looking for someone to set up my a.i agents to run my real estate wholesale business. i already have the platform just need yo to set up my triggers and systems. If you can do this let me know! Thanks

Malaki I.

We're looking for a senior-level GoHighLevel specialist who can own our CRM, automations, and AI systems end-to-end - not someone who simply executes instructions.

We need independent thinking and ownership.

Non-Negotiable Requirements

Only apply if you have ALL of these:
1. 2+ years hands-on GHL experience in live operational businesses. You have done full builds (automations, funnels, pipelines, reporting, membership areas, GoKollab) for actual businesses

2. Must be fluent in English, have a reliable internet connection, and be available during hours that overlap with Australian business hours (AEST/AEDT)

3. Proven AI bot builds: conversational flows, decision logic, escalation rules

4. You troubleshoot fast and understand conditional logic, triggers, and integrations

5. You can clearly explain what you built, why, and the results it produced

6. Daily Asana updates, Looms, reliable availability. You never need chasing

7. Comfortable with time-tracking software for team accountability and capacity planning

How to Apply

These steps are your first test. Applications that don't follow instructions will not be reviewed.

1. Enneagram Test: Take the free test at truity.com/test/enneagram-personality-test. Attach a screenshot of your full result including your "superpowers."
Please note: You do NOT need to pay to take the test or get the result.

2. Two-Minute Video: Record a max 2-min video (YouTube, Loom, or Drive) introducing yourself and sharing whether your Enneagram result feels accurate and why.

3. Location & Compensation: Include your location, time zone, and target annual salary.

If you've read this far and thought "That's me, and I can prove it", we're excited to see what you've built.

• Our core value is simple: Set The Standard. Walk us through an experience you had where you believe you have set the standard.

We're building Decked - the first pre-construction sales platform for residential contractors. There is no software that exists for the moment between a homeowner's first inquiry and a signed contract. Every tool in construction (Procore, Jobber, Buildertrend) starts after the contract is signed. Decked operates before it.

Here's what the product does:

A homeowner goes to a contractor's Decked link. They answer 8 blocks of questions through an AI conversation - scope, budget, timeline, materials, photos of their space. Before they finish, the contractor receives a complete lead brief via SMS and email: scope, budget, timeline, three AI-generated visualizations of the homeowner's actual space with their chosen materials applied, a preliminary cost estimate, and a permit flag. The contractor walks into the first meeting knowing everything. The homeowner walks in having already committed to a direction.

That's Phase 1. The platform expands from there into a full pre-construction operating layer - proposal builder, CRM pipeline, Smart Outreach automation, agency white-label, and eventually permit filing, material sourcing, and subcontractor matching.

We're pre-launch, actively raising a pre-seed round, and ready to build.

Want to go deeper before applying? We've documented everything - the full product spec, the roadmap, the business model, the vision. Read it here: https://decked.notion.site/


WHAT THIS ROLE COVERS
This role covers two things. If you can do both well, we'd rather work with one person than split it. If you're a strong product engineer but not a website person, tell us that and apply for the product build only.

Part 1 - Marketing Website
Part 2 - Full Product Build

There's also a third component: technical strategy and advice. We're founders, not engineers. We need someone who does more than write code - someone who thinks with us. Architecture decisions, API choices, scalability tradeoffs, what to build now vs. later, what to use off the shelf vs. build in house. We want a technical partner who tells us when something in the spec doesn't make sense and proposes something better. If you only want to execute tickets, this isn't the right fit.


PART 1 - MARKETING WEBSITE
We need a clean, fast, credible marketing site live as soon as possible. Not perfect - live. While the product is being built, investors and contractors need somewhere to land. Five pages, fast turnaround, done well.

Pages needed:
- Home - above-the-fold hook, what Decked does in plain language, who it's for, social proof section (placeholder for now), clear CTA
- For Contractors - what they get, how it works, why it changes the first meeting, pricing overview
- For Agencies - the white-label pitch, portfolio management, why one agency deal brings 10-20 contractors onto the platform
- About - founding team, what we're building and why
- Contact / Get Started - simple form or calendar embed

Requirements:
- Platform is open - Framer, Webflow, whatever you're fastest and best in. If you have a strong SaaS template you can adapt quickly, use it. Speed matters.
- Mobile-first. Flawless on iPhone Safari and Android Chrome.
- Clean, premium feel. Think fintech meets proptech. Not a template that looks like a template.
- Fast load times. No bloat.
- Basic analytics integration (we'll provide the tracking ID)
- Copy provided by us. You handle design and build.
- Domain connection and go-live included.

On brand: we're flexible on color direction - we trust your eye. The tone is confident, clean, and professional. Think somewhere between Stripe and Linear. Not loud. Not generic. We'll share our logo and any existing assets at kickoff. If you see a direction that works better, bring it.

Timeline for the website: under 2 weeks from kickoff to live.


PART 2 - FULL PRODUCT BUILD
Three portals on one shared data layer.

Homeowner Portal (decked.com)
- Mobile-first, no account required
- 8-block AI conversational intake, iMessage-style, one primary action per screen
- Progressive disclosure - homeowner never sees what's coming next
- English and Spanish throughout
- Block 01 submission fires lead brief to contractor in real time, under 30 seconds
- Block 08: photo upload, up to 8 room photos, floor plan, inspiration images
- Post-consultation: three rendered visualization views of the homeowner's space with their material selections applied
- Three-change flow: homeowner can request up to 3 revisions conversationally, each re-renders
- Full change log preserved and delivered to contractor

Contractor Portal (pro.decked.com)
- Desktop-first, responsive
- Business profile setup, 7-minute onboarding target
- Past jobs entry for estimate calibration
- Live lead feed with seriousness signal (High/Medium/Low based on block completion)
- Lead detail view: full brief, all three visualizations, flag and suggestion log (Sold/Verify/Confirm per material decision), preliminary estimate hidden from homeowner by default, permit flag
- Pipeline: 5 stages (Contacted / Site Visit Scheduled / Proposal Sent / Won / Lost), list and kanban toggle, filterable
- Incomplete leads pinned at top with one-tap call
- Edit Mode: contractor corrects dimensions, materials, scope after the site visit; triggers automatic re-render and estimate recalculation; original homeowner inputs preserved in version history
- Estimate visibility toggle per lead: Hidden / Include in Proposal / Share as Range
- AI assistant that reads live account data and gives lead-specific advice and pipeline coaching
- Automatic follow-up SMS for incomplete leads, 24 hours, through original channel
- Post-call auto-send fires 2 minutes after contractor hangs up
- SMS delivery must route as iMessage (Blue Text) where possible. Standard SMS has high spam filter rates and low open rates. All outbound SMS - follow-ups, consultation links, post-call auto-sends - need to go out as Blue Text to ensure delivery and trust. This is a technical requirement. Tell us your approach in your proposal.

Estimate Engine and Material Sourcing
- RSMeans baseline plus contractor's own past job calibration
- Confidence scoring: High/Medium/Low
- Recalculates automatically when contractor uses Edit Mode
- Material sourcing: the homeowner's material selections from the consultation generate an itemized material list the contractor can access directly from the lead brief. Pre-populated supplier links with the correct material pre-selected. One-tap access from lead detail view. The contractor should not have to reconstruct this list manually.

Smart Outreach (MVP baseline)
- Contractor's unique consultation link live immediately on profile setup
- Post-call auto-send toggle via Blue Text SMS
- Manual link share from lead card

Timeline for the product: 3-4 months from contract to first beta users.


ON DESIGN
We care about design. A lot. This is not a product where ugly is acceptable because it's B2B. The homeowner experience especially needs to feel premium - it's consumer-facing and needs to feel like a fintech app, not a contractor tool. The contractor dashboard needs to be clean and focused, not cluttered.

We're not asking you to be a full-time designer. But we need someone who has taste, has worked with designers or can direct design, and won't ship something that looks like a default template. If you have a strong design sensibility yourself, tell us. If you'd bring in a UI/UX collaborator, tell us how you'd handle that.

Describe how you think about product design in your proposal. It matters to us.


TECH
We don't have a locked stack. Tell us what you'd use and why. General direction:

- Frontend: React or Next.js
- Backend: Node.js or Python
- Database: PostgreSQL
- Hosting: Vercel and Railway or similar
- AI/LLM: Anthropic Claude API or OpenAI
- Visualization: this is the most technically interesting part - tell us your approach

The visualization engine - rendering homeowner-uploaded room photos with their chosen material selections applied - is the hardest technical problem in the MVP. We want to know how you'd approach it, what services or models you'd evaluate, what the tradeoffs are, and what a realistic quality bar looks like for Phase 1 vs. later. There's no single right answer. We're looking for someone who has thought it through.

We're also open to your input on anything in the spec you'd approach differently. If something should be a third-party service instead of a custom build, a different data model, a phased approach we haven't considered - we want to hear it. That's part of what we're hiring for.


WHAT WE'RE LOOKING FOR
Required:

- Shipped multi-sided SaaS products, not just CRUD apps or single-sided tools
- Strong full-stack, frontend and backend, not just one
- Experience integrating LLM APIs in production
- Experience with image processing, rendering, or computer vision
- Can own technical decisions, not just execute them
- Strong communicator

Strongly preferred:

- Real-time data experience
- Mobile-first consumer-facing frontend work
- Design sensibility or experience directing designers
- Experience advising early-stage founders on technical decisions

Please answer these in your proposal:

- How would you approach the visualization engine? What tech, what process, what are the hard parts, what's realistic for MVP vs. later?
- What's a multi-sided platform you've shipped? What was the hardest technical decision you made on it?
- What stack would you use and why?
- How do you think about design? Do you have a design sensibility yourself or do you work with designers?
- Can you take on both the website and the product build, or just one?

Generic proposals will not be read.


HOW WE WORK
Ship something live as fast as possible. It doesn't need to be finished. While you're still building, we want something in front of contractors and investors. A functional homeowner consultation flow with a basic contractor brief view is more valuable to us live at week 6 than a polished complete product at month 4. We'll tell you what to prioritize for the first live version.

We expect things to change. As we talk to contractors, as we raise, as the product gets in front of real users - things will shift. Scope will get refined. Priorities will move. We need someone who treats that as normal, not as a problem. Build to iterate.

Communication over everything. Weekly updates at minimum. Tell us what you built, what's blocked, what you're deciding and why.


COMPENSATION
Posted budget: $15,000. This is a starting point, not a ceiling.

For the right person, total compensation is up to $40,000, paid in milestones. We're posting $15K because of how Upwork surfaces posts by budget - the real number is meaningfully higher and negotiable based on scope ownership and what you bring beyond the build.

Example milestone structure (negotiable):

- M1: Website live, architecture locked, homeowner consultation flow functional - 20%
- M2: Visualization engine working, lead brief fires in real time - 25%
- M3: Contractor dashboard complete with Edit Mode, pipeline, estimate engine, material sourcing - 25%
- M4: Smart Outreach basics, Blue Text delivery confirmed, QA pass - 20%
- M5: Beta-ready, bugs resolved, go-live - 10%


THE BIGGER OPPORTUNITY
We're a pre-seed startup actively raising. We're not hiring a full-time CTO right now - that hire happens post-raise. But this engagement is how we find that person.

If you're the right person on this project:

- Ongoing contract work through Phase 2 and 3 of the roadmap (analytics, proposal builder with e-signature, CRM integrations, scheduling, guided photo capture)
- Equity as a founding engineer - this is a real conversation for the right person, not a teaser line
- First consideration for the CTO role when we raise - we want to hire from the relationship, not from cold outreach

We're building something real in a $498B market with no direct software competition in our lane. If you want to be part of a company from day zero, not just a freelance engagement, this is the conversation to have.

We are looking to hire an experienced AI developer / automation specialist who can help us build AI agents, internal tools, and client-facing AI solutions for our digital agency.

We need someone who truly understands AI technology, not just someone who has experimented with ChatGPT prompts. The right person should have hands-on experience building AI workflows, AI agents, automations, integrations, and practical business tools.

About Us

We are a digital agency working with WordPress, WooCommerce, custom websites, SEO, Google Ads, hosting, email marketing, and business automation. We are expanding our services into AI-powered solutions for both our internal operations and our clients.

We want someone who can help us design, build, test, and improve AI systems that actually solve business problems.

Responsibilities

You will help with:

Building AI agents and AI-powered workflows
Creating automations for internal agency tasks
Connecting AI tools with websites, CRMs, forms, email, project management tools, and databases
Helping us understand the best AI technology stacks for different use cases
Building prototypes and production-ready AI tools
Creating AI assistants for customer service, sales, marketing, reporting, and operations
Working with APIs, webhooks, and third-party integrations
Helping document processes and teach our team how to use the systems
Advising us on the best tools, platforms, and frameworks to use
Technology Experience We Are Looking For

The ideal candidate should have strong experience with some or most of the following:

OpenAI API / ChatGPT / Assistants / GPT models
AI agents and multi-step AI workflows
LangChain, LangGraph, CrewAI, AutoGen, n8n, Make, Zapier, or similar tools
Vector databases such as Pinecone, Supabase, Weaviate, Qdrant, or similar
RAG systems and knowledge-base AI assistants
API integrations and webhook automation
WordPress and WooCommerce integrations
CRM, email marketing, and project management integrations
JavaScript, Python, Node.js, or similar development skills
Cloud platforms such as AWS, DigitalOcean, Cloudflare, or similar
Databases, structured data, and document processing
AI-powered content generation, reporting, and business automation
Important Requirements

We are looking for someone who:

Has real-world experience building AI tools and agents
Can explain technical concepts clearly
Can recommend the right stack instead of just using trendy tools
Understands business use cases, not just code
Can work with an existing agency team
Can build scalable, secure, and maintainable solutions
Is comfortable experimenting, testing, and improving systems over time
Can document what they build
Has strong communication skills
Example Projects

Some projects we may want to build include:

AI assistant for managing website maintenance requests
AI-powered SEO reporting assistant
AI customer service chatbot trained on a company’s website and documents
AI lead qualification agent
Internal AI project manager assistant
AI content generation workflow for blogs, emails, and landing pages
AI agent that connects with WordPress, forms, CRMs, and email
AI-powered client onboarding workflow
AI assistant for analyzing Google Analytics, Search Console, and ad performance data
Please Include in Your Application

When applying, please answer the following:

What AI agents or AI automation tools have you built before?
What AI technology stack do you recommend for a digital agency and why?
Do you have experience with WordPress, WooCommerce, APIs, and webhooks?
What tools do you prefer: LangChain, CrewAI, n8n, Make, Zapier, custom Python/Node.js, or something else?
Please share examples of past AI projects, workflows, or demos.
How would you approach building an AI assistant for a digital agency?
Type of Work

This can start as a project-based role, but we are open to an ongoing long-term relationship if it is a good fit.

We are looking for someone who can become our go-to AI technology partner as we expand our agency into AI services.

We are a California-based HVAC and home performance contractor seeking a freelancer/consultant to design, build, and implement AI-driven operations. The role involves optimizing processes, enhancing efficiency, and integrating AI solutions to improve our services. The ideal candidate will have experience in operations management and AI, with a focus on HVAC or home performance.

We are building out a GoHighLevel implementation for a multi-location photography operation in Hawaii with 19 locations across 3 brands. We need an experienced GHL specialist who can work alongside our IT consultant to build and configure the platform from the ground up — calendars, automation workflows, AI bots, and integrations. This is a substantial build and we are looking for someone with real multi-location and automation experience, not just basic funnel building.

Our Stack
GoHighLevel (agency account — Pacific Dream Photography sub-account)
Square for payments and photographer commission tracking
ShootProof for photo gallery delivery
Google Workspace for email

What We're Looking For
Proven GHL experience — please share examples of multi-location builds, not just simple funnels or lead gen setups
Strong workflow and automation builder — we need complex multi-step sequences not basic confirmations
AI bot configuration experience — specifically GHL's AI conversation flows and lead qualification
Clear communicator who can work alongside our IT consultant and explain what they're building
Available for ongoing support after the initial build — this is not a one-time project

Engagement
This starts as a project-based engagement with the expectation of ongoing support as we expand and add locations. We are looking for someone who wants to build a long-term working relationship, not just complete a ticket and disappear.
Please include in your proposal:
Examples of multi-location GHL builds you have completed
Your experience with AI bot and automation workflow configuration
Your hourly rate and estimated hours for a build of this scope
Your availability and timezone

About the Project
Pacific Dream Photography operates across Maui, Oahu, the Big Island, and Kauai serving resort and hotel guests through concierge referrals, hotel QR codes, Yelp, Google, and direct website inquiries. The goal is to reduce manual admin work, respond to leads instantly 24/7, and build a fully automated booking and follow-up system that scales as the business grows.

• Describe your experience with Multi-Location Go High Level implementations.
• Describe a complex automation workflow you've built in GHL including AI and multi-step sequences.