An Enterprise-Grade, Multi-Agent AI Platform for Autonomous Career Management
This platform is a state-of-the-art AI Career Copilot designed to demonstrate advanced software engineering, distributed systems, and AI agent orchestration. It autonomously discovers, scores, and manages job opportunities using a hyper-personalized professional profile.
Built with an uncompromising focus on scalability, security, and Twelve-Factor App principles, this project reflects senior-level engineering practices ready for production environments.
- 🧠 Autonomous Multi-Agent Swarm: A decentralized architecture of 8 specialized AI agents (Scraping, Ranking, RAG, Q&A, Cover Letter, Orchestrator, Observability, and Security) working in tandem via Temporal workflows.
- 🎯 Semantic Matching (RAG): Uses
pgvectorand deep semantic search to rank job descriptions against candidate CVs with high precision, removing the noise of traditional keyword matching. - 🛡️ Enterprise Security: Built-in OWASP hardening, Supabase JWT authentication, Role-Based Access Control (RBAC), Row-Level Security (RLS), and active prompt injection defenses.
- 📊 Deep Observability: Comprehensive telemetry with OpenTelemetry, Prometheus, Loki, Grafana, and Sentry for real-time monitoring of AI hallucinations, token budgets, and system latency.
- ☁️ Cloud-Native Infrastructure: Infrastructure as Code (IaC) via Terraform, supporting multi-cloud deployments to Azure Container Apps and AWS ECS Fargate, fully containerized with Docker.
- Framework: Next.js 16 (App Router, Static Export), React 19
- Styling: Tailwind CSS v4, Modern Flexbox & Grid, Responsive Design
- State Management: TanStack Query (React Query) featuring high-performance Keyset (Cursor-based) Pagination capable of scaling to millions of rows.
- Language & Framework: Python 3.14, FastAPI (Async),
uvpackage manager - AI Engine: Multi-Agent Architecture, RAG, Prompt Engineering, Strict XML Prompt Contracts
- Web Scraping: Playwright (Headless browser automation), BeautifulSoup
- Orchestration: Temporal for distributed, fault-tolerant workflow management and dead-letter queues.
- Primary Database: PostgreSQL (Supabase) with
asyncpg - Vector Search:
pgvectorfor localized AI embeddings - Caching & Rate Limiting: Redis (
aioredis) with sliding window algorithms - Schema Migrations: Alembic
- CI/CD: GitHub Actions (Linting, Ragas/DeepEval Regression testing, Terraform pipelines)
- Containerization: Docker multi-stage builds, Nginx reverse proxy, Supervisord
- Security: Cosign (Docker Image Signing), Bandit, Trivy
The system is powered by a decentralized swarm of AI agents, each governed by strict XML contracts and isolated responsibilities:
| Agent | Responsibility |
|---|---|
| Scraping Agents | Navigates volatile session URLs on LinkedIn/JobServe to extract and standardize data. |
| Ranking Agent | Scores jobs against candidate CVs using cross-encoder confidence intervals. |
| RAG Agent | Contextualizes job descriptions with candidate history via vector search. |
| Q&A Agent | Provides grounded, hallucination-free interactive answers about job postings. |
| Cover Letter Agent | Dynamically generates ATS-optimized, personalized cover letters. |
| Orchestrator Agent | Manages long-running workflows, circuit breakers, and retries via Temporal. |
| Observability Agent | Monitors AI schema conformance, tracks token budgets, and alerts on anomalies. |
| Security Agent | Sanitizes user inputs and actively guards against prompt injections. |
- Docker and Docker Compose
- Node.js 22+
- Python 3.14+ and
uvpackage manager
The entire stack is orchestrated via Docker Compose for a seamless setup.
Note: An
.envfile must exist in the root directory before running Docker. Copy.env.exampleto.envand fill in your API keys and configuration.
cp .env.example .env
# Edit .env with your credentials
docker compose up -d --buildThe unified application is now accessible at: http://localhost
1. Start the Backend (FastAPI)
cd backend
uv sync
uv run --project backend uvicorn backend.main:app --reload --port 8000API Docs: http://localhost:8000/api/docs
2. Start the Frontend (Next.js)
cd frontend
npm install
npm run devDashboard: http://localhost:3000
This project strictly adheres to Twelve-Factor App (2026) principles, demonstrating a mature engineering mindset:
- Documentation Driven: Every module and agent contains localized
AGENT.mdguidelines outlining input/output contracts. - Reliability First: Integrated DIFA (Discover, Isolate, Fix, Assess) framework and ReAct loop execution protocols.
- Test-Driven AI: Prompt regression testing integrated directly into the CI pipeline using DeepEval and Ragas.
- Zero Magic: Explicit over implicit. No "pseudo-code" allowed in production—every feature is fully implemented and tested.