🚀 Forward Deployed AI Engineer @ BCG X
🎓 MVA Master's Graduate @ ENS Paris-Saclay · Ingénieur @ IMT Atlantique
💻 Agentic AI · Machine Learning · Deep Learning · Computer Vision · LLMs
🚀 Forward Deployed AI Engineer at BCG X, helping organizations design, build, and deploy high-impact AI solutions — from LLM orchestration and agentic pipelines to robust backends and human-in-the-loop interfaces. I work directly with clients to frame the right problem, then build the right solution, from prototype to production.
🎓 Graduated from a Master's degree in Mathematics, Vision, and Learning (MVA) at ENS Paris-Saclay, one of the top AI programs in Europe, and hold an engineering degree from IMT Atlantique.
🔬 I'm particularly drawn to complex, frontier problems:
- 🤖 Agentic AI and multi-agent orchestration
- 🧠 Large Language Models, Transformers, Mamba
- 🖼️ Computer Vision and Embedded AI
- ⚙️ Distributed Training and Optimization
- 📊 Building robust ML systems that scale
Published at EUSIPCO 2024, with a submission to CVPR 2026. Curious by nature, pragmatic by conviction.
-
🚀 Forward Deployed AI Engineer @ BCG X (2026 - Present) Designing and deploying agentic AI platforms for enterprise clients — LLM orchestration, agentic pipelines, and production-grade human-in-the-loop systems.
-
🛰️ AI Research Intern @ Thales (2025) Implemented Mamba-based architectures for high-resolution semantic segmentation; comparative study against Transformers (ViT, Swin); optimized for embedded inference on NVIDIA Orin AGX with TensorRT acceleration.
-
🤖 Data Scientist LLM @ Polynom (2024) Fine-tuned open LLMs for production use cases (LoRA, PEFT); built and optimized RAG pipelines (embeddings, retriever-ranker architecture, end-to-end evaluation).
-
🔬 Research Project @ IMT Atlantique (BRAIn team) (2023-2024) Authored a research paper on few-shot image classification via foundation-model-based unsupervised segmentation.
-
🖥️ System Administrator @ ResEl (2021-2024) Managed the student-run ISP network across IMT Atlantique campuses (700+ rooms); built VM management tooling for datacenter energy optimization.
| ⭐ Project | 🚀 Description | 🏷️ Tags |
|---|---|---|
| HydraScale | Ultra-Scale LLM Training Engine. A from-scratch PyTorch framework to benchmark parallelism strategies (DDP, FSDP/ZeRO). Implements advanced data loading, hardware-agnostic training, and profiler-based bottleneck analysis. | Distributed Training, LLM, PyTorch, HPC, FSDP, ZeRO |
| Yubu Code | AgentOps Replay System - AI agent transparency, compliance & debugging | AI Agents, LangChain, FastAPI, Next.js |
| FICUS | Unsupervised segmentation for few-shot classification | Computer Vision, Few-Shot, Research |
| LongGPT | Implementation of LongNet to scale sequence length >1B tokens on top of nanoGPT | LLM, Long Context, Transformer |
🛠 Explore more in the Repositories tab »
🇫🇷 French (Native) · 🇬🇧 English (Full Professional) · 🇪🇸 Spanish (Limited Working) · 🇯🇵 Japanese (Elementary)
"The task is, not so much to see what no one has seen yet; but to think what nobody has thought yet, about what everybody sees." - Arthur Schopenhauer