I architect production-grade multi-agent LLM pipelines, RAG systems, and serverless AI backends — then ship them to prod with the discipline of an athlete training 10-15h a week. Currently building cross-cloud AI solutions at Formalis.IA (AWS + GCP), automating complex legal document workflows at scale.
- 🔧 Designing fault-tolerant agentic architectures (LangGraph, ReAct) with deterministic supervisors
- 📚 Researching advanced RAG evaluation and retrieval ranking for domain-specific knowledge bases
- 🤝 Open to collaborating on multi-agent systems, MLOps pipelines, and open-source LLM tooling
Architecture & Research
Clean Architecture DDD TDD Repository/Factory/Ports & Adapters Bayesian Inference Markov Chains Information Theory
|
🛡️ Agentic Credit & Geopolitical Risk Multi-agent LLM pipeline → CRO-level risk reports in <60s 3 autonomous agents (ReAct + LangGraph) with Redis checkpointing, ChromaDB RAG, and full LangSmith observability.
|
📡 Multimodal Alpha Signal Extractor Fine-tuned VLM on candlestick charts + NLP sentiment scoring for live trading signals Custom pipeline combining vision-language models with real-time market data. |
|
Sustainable ledger system integrating localized predictive analytics with robust backend infrastructure. |
⚡ PCBA Cost Prediction — Schneider Electric End-to-end ML pipeline with Stacking Ensemble (XGBoost, RF, GBM, Ridge)
|
- 🎓 Specialized Master in Big Data — GEM & Ensimag, Grenoble (2025–2026)
- 🎓 Engineering (Electrical & Energy) — ESME, Lille (2022–2025)
- 🎓 BA1 Mechanical Engineering — EPFL, Lausanne (2021–2022)
- 📜 Stanford Machine Learning Specialization (2025)
- 📜 Stanford Deep Learning Specialization (2025)
Same discipline. Different track.
| 🏃 Athletics | 800m 1'55"77 · 1500m 3'58 · 3000m 8'48 · 5km 15'19 · 10km 33'20 · HM 1:14' · M 2:46' |
| 🧗 Rock Climbing | Since age 4 — translating route-reading into system design |
| 🏉 Rugby | 9 years competitive · international matches across Asia |
| 🌏 Expatriate | 3+ years in Bangkok, Thailand · fluent FR / EN / ES |