Computer Engineering student @ EPITA (🇫🇷) | Chapter Lead @ GDG EPITA | Independent ML Researcher
I'm a builder, entrepreneur, and community leader — equally energized by low-level systems programming and high-level product strategy. I lead the Google Developer Group (GDG) EPITA community (workshops, tech talks, hackathons), co-founded and led engineering at ChainPilot as CTO, and I'm currently writing an independent first-author research paper on how models understand user intent.
"Turning complex problems into delightful, user-centred solutions."
💼 Exploring — empirical AI safety research roles. Interests: mechanistic interpretability, scalable oversight, adversarial robustness, AI control.
| 🧠 AI & LLM agents | Multimodal reasoning, multi-agent orchestration, and tool use (MCP) |
| 📄 Research | First-author preprint on coverage-driven user intent understanding, targeting AAAI 2027 |
| 🌐 Full-stack & cloud | Scalable back-ends, databases, CI/CD, and DevOps pipelines |
| 📱 Mobile & spatial computing | SwiftUI, Vision Pro apps, and AR/VR experiences |
| ⚡ Systems & algorithms | Low-level C/C++, data structures, optimisation, and performance |
Outside of code, I mentor peers, organize hackathons, write technical deep-dives, and nurture a healthy obsession with specialty coffee ☕.
| Project | Stack | Brief Description |
|---|---|---|
| Telos 🔗 repo · Devpost |
Python / Next.js / FastAPI | Turns one natural-language instruction into a deployed product via a self-healing multi-agent build loop. Built at Hackeurope, now a validation case study in my research |
| ChainPilot 🔗 site |
Python / TypeScript / Next.js | Financial logic layer for the agentic economy — blocks unsafe on-chain actions before execution. Co-founded and led engineering as CTO (Nov 2025–2026) |
| RadAssist 🔗 repo |
Python / Jupyter | Agentic workflow drafting context-aware radiology reports and tracking lesion progression, aimed at cutting radiologist burnout |
| InVision SF 🔗 repo |
SwiftUI / RealityKit | Vision Pro app that places an interactive 3‑D map of San Francisco events on any tabletop, built with the SF Mayor's Office |
| OCR Word Search Solver 🔗 repo |
C / GTK4 | Solves word-search puzzles from a photo: image binarization, a hand-written neural net for character recognition, and a grid solver |
| English–Korean Translator 🔗 private |
Python / TensorFlow | Custom seq‑to‑seq model (attention + beam search) trained on 100k+ sentence pairs |
📄 A public repo for the research paper (Coverage-Driven User Intent Understanding, targeting AAAI 2027) is coming once it's ready to share.
A snapshot of what I'm sharing — see more on my Medium profile:
- Reimagining Programming Education: Why We Built BitLyne – Insights into building a gamified coding-education app and rethinking how people learn programming. [Medium]
- From Paris to San Francisco: What I Learned Studying Computer Science in Two Countries – Reflections on studying CS in France and the U.S., and how cultural/academic contrasts shaped my path. [Medium]
- Building While Learning: How College Projects Shaped My Engineering Playbook – Lessons from turning coursework into production-grade code. [Medium]
- MathCam — Solving Math Through the Lens – Turning camera snapshots into step-by-step solutions with on-device AI. [Medium]
Building bridges between people and technology — one commit at a time.