Status: early development (Week 0/8) — not yet functional.
An agentic AI system that matches a candidate's CV against AI/ML job postings in Germany, explains why a job fits or doesn't, and ranks recommendations — built to demonstrate real engineering judgment (deterministic logic vs. LLM reasoning, RAG, tool calling, evaluation, and deployment), not just a LangChain wrapper around an API call.
This repo is being built incrementally and in the open over 8 weeks. Each
tagged version (v0.1, v0.2, ...) adds one real, tested capability — see
Progress below.
FastAPI → LangGraph agent → tools (job search / CV retrieval / match scoring) → PostgreSQL + pgvector → evaluated with Ragas, traced with Langfuse → Docker → Azure.
Full design rationale (why pgvector over a dedicated vector DB, why
LangGraph, why Azure, etc.) lives in docs/architecture.md (coming in
Week 5-6).
| Version | Capability | Status |
|---|---|---|
| v0.1 | Job description → structured JSON extraction | 🔲 not started |
| v0.3 | CV ingestion + RAG retrieval | 🔲 not started |
| v0.5 | Tool calling / agent loop | 🔲 not started |
| v0.7 | REST API + MCP server | 🔲 not started |
| v1.0 | Deployed on Azure | 🔲 not started |
(To be written once there's something to set up.)
Built as a portfolio project targeting AI Engineer / Generative AI Engineer /
Agentic AI roles in Germany. Every design decision is documented and
defensible in an interview — see docs/ as it fills in.
This project is licensed under the MIT License.
Copyright © 2026 Mehran Pashaei.