Work-study engineer (R&D, AI), CAPFI
Jan 2026 – present · Paris
Lead developer of Cerebro, an internal tool that matches consultant profiles with client tenders using LLMs.
Designed and put into production an asynchronous AI pipeline for extraction and scoring for tender × profile pairs, processing several hundred records via reliable batch management.
Calibrated the scoring prompt (temperature, required or desired skills, three sub-scores) and added short justifications so users can verify each score.
Built an annotated benchmark dataset to quantify and reduce the false positive and false negative rates of the scoring system.
Automated the generation of daily reports for a team of ten salespeople and secured access to the API and web interface via SSO authentication. Stack: Python, FastAPI, Pydantic, SQLite, Docker, GitLab CI, Gemini.