Computer-science PhD specialised in machine learning, engineering AI products for healthcare & business.
AI researcher specialised in healthcare — multimodal learning, clinical risk, generative AI. I also ship practical AI products for hospitals and small businesses.
I'm an AI & Machine Learning engineer with a PhD in computer science (Nov 2023), specialised in healthcare. My doctoral research, run with Amiens University Hospital, focused on motion analysis with AI — applied notably to Parkinson's disease.
I'm currently a post-doctoral researcher at the University of Picardie Jules Verne on the RAUC project: an optimised care pathway for digestive surgical emergencies, using AI to predict post-operative complications. Before that, I led the AI & Data Science workstream at Capgemini Engineering on a generative-AI medical assistant, coordinating a team of 15–20.
In parallel I build practical AI products for healthcare and small businesses, with deep expertise in LLM systems, multimodal RAG, computer vision (medical imaging), explainable AI and production ML.
Products & applications — in use or in active development.
Full-stack ERP for French artisans & small businesses, in late-stage development: quotes, invoices, credit notes, deposits, e-invoicing (Factur-X), Stripe billing, OCR scanning, voice dictation (Whisper) and a conversational AI assistant. GDPR audit, in-house 'Atelier' design system.
Healthcare POC generating structured French medical prescriptions from voice dictation (local whisper.cpp) via a Claude agent driving an MCP tool server — regulated PERFADOM / OAH home-care forms, RBAC and drug-interaction checks. 100% synthetic data; archived — it led to Ordo, a narrower, market-focused successor.
Web app I built for a radiologist at Amiens University Hospital to generate the department's monthly schedule — 5 specialties, 6-week rotation, manual locks with automatic regeneration. Fully client-side (React + Vite), persisted in localStorage.
Cross-platform Flutter app with a Supabase back-end: offline-first logging (Drift/SQLite) with cloud sync, real-time shared sessions, and a rule-based weekly progression engine. In closed beta on Google Play — AI-assisted plan generation on the roadmap.
Small open-source experiments — Python code where an agent learns to play or evolve from scratch.
100 agents learn Flappy Bird from scratch through NEAT neuro-evolution; one beats the human record in ~30 generations.
GitHub ↗Shooter vs. keeper trained by self-play, converging to the game-theoretic Nash equilibrium (mixed strategy).
GitHub ↗Mini-agents learn 1v1 and team football by reinforcement learning — shooting, dribbling and saving emerge on their own.
GitHub ↗A Deep Q-Network agent learns to play Pac-Man from the raw game state.
GitHub ↗An RL agent plays Space Invaders with a live visualisation of the neural network as it fires.
GitHub ↗Snake mastered by Deep Q-Learning, with a multi-agent split-screen comparison of strategies.
GitHub ↗Artificial-life sim: cells driven by tiny neural networks that reproduce, mutate and evolve by natural selection.
GitHub ↗GenAI medical-assistant demonstrator with multimodal processing: OCR for blood tests, Whisper STT, fine-tuned LLMs for clinical summarisation, multimodal RAG.
Transformer segmentation (UNetR, SwinUNet) for earlier preeclampsia detection on 3D Doppler images. Submitted to CMIG 2025 (co-author).
Motion modeling for Parkinson's disease and activity recognition. Deep learning (RNN, LSTM, CNN), genetic algorithms, interpretable matrix representations. 3 publications.
I'm always up for discussing AI & ML opportunities — especially in healthcare and applied products. Reach out.