Spaces:
Sleeping
Sleeping
| AI engineering experience: | |
| Founder & AI Engineer, Plaidoyer.ai, London, 01/2026 to present | |
| I built a software platform that helps French lawyers quickly find legal information, manage case documents | |
| securely, and use AI to assist with research and client communication, ensuring full privacy compliance. | |
| Responsibilities: | |
| - Developed hybrid retrieval pipeline using PostgreSQL pgvector and HNSW indexing for 2.9M+ French legal | |
| documents | |
| - Implemented semantic search with BGE-M3 embeddings and full-text search capabilities | |
| - Integrated cross-encoder reranking and HyDE query expansion for improved search accuracy | |
| - Designed LLM-based query requalification with conversational clarification loop | |
| - Built RAG assistant for document chat with strict tenant isolation and prompt-injection defenses | |
| - Engineered data ingestion pipeline from LegiFrance/DILA APIs with rate-limited batch extraction | |
| - Architected compliance-first LLM system ensuring EU-only inference and zero data retention | |
| - Developed backend with FastAPI, SQLAlchemy, and Alembic + frontend using React and TypeScript | |
| - Deployed and managed infrastructure on Hetzner and AWS platforms (Python, AWS, Docker, Linux, Alembic, | |
| HNSW, TypeScript, SQLAlchemy, Hetzner, LLM, FastAPI, React, Amazon S3, JWT, Mistral AI, Rag, BGE-M3, | |
| PostgreSQL, Pgvector) |