I build products that run in production, experiment with AI models, and secure the systems I ship.
About
With an MSc in Cybersecurity (Guardia, 2025) and a BSc in Artificial Intelligence (Simplon, 2023), I combine machine learning, full-stack development and information security. Three years of work-study across finance (Amundi, Crédit Agricole), an environmental nonprofit (Surfrider Foundation) and a large-account project (Axiome Solutions) taught me to ship as a team across varied stacks - from a YOLO model in production to React/TypeScript features delivered each sprint.
In my own time, I take side projects all the way to production: a native iOS news app shipped end-to-end (app + backend API), and a Rocket League agent trained with Deep Reinforcement Learning (8.17 billion steps, Platinum rank). I'm looking for a full-time AI or data engineering role - open to relocation across France and internationally (V.I.E).
Products that run in production, from backend to mobile.
Vertical-format RSS news reader - native iOS app + backend API, shipped end-to-end.
Native SwiftUI iOS app consuming a FastAPI backend hosted on a Linux VPS. Server side: aggregation of dozens of RSS feeds, 15-min cache per category, filtering of stale articles. App side: infinite vertical scroll, user-reorderable categories, offline mode with local cache, custom pull-to-refresh.
Game server backend (Minecraft Paper) with custom-built Java plugins.
Terminal-first private network linking 3 remote machines over a VPN mesh.
Two fictional sites built to practice editorial design and micro-interactions. No real product behind them - just the urge to push the frontend in two opposite directions on the same stack (Next.js 16, Tailwind v4, Motion).
UnpublishedExercise - synthwave record shop (Miami Outrun)
Fictional e-commerce exercise: Stripe (test mode), EN/FR i18n, admin back-office with HMAC cookie auth, full checkout flow.
UnpublishedExercise - studio showcase (1944 dark)
Cinematic one-page showcase exercise. Dark mood, orange gradient, anime.js animations. No real studio.
Research and experiments - Deep RL, models, distillation.
Rocket League bot trained with Deep RL.
A Rocket League bot trained with Deep Reinforcement Learning, from scratch, with no human supervision - only rewards and 8.17 billion simulated steps. Custom training pipeline (PPO on RLGym v2 + RocketSim) on cloud GPU (RunPod RTX 4090), tested in real conditions via RLBot. Architecture distilled from an MLP teacher (7.5M parameters) to a lighter EARL Transformer (3.3M). Platinum rank reached, end goal Grand Champion (top 15%).
Master's thesis on deepfake detection with CNNs. Full pipeline: frame extraction, face detection and cropping, classification via fine-tuned Xception. Per-video split to avoid identity leakage, unified train/val/test preprocessing.
Running my own systems, hardening them, monitoring them.
I host several services in production on a personal VPS. Continuous hardening, isolated secrets, reduced attack surface, active monitoring - the minimum for a stack to last over time without constant supervision.
Guardia · 2023 - 2025 · completed
Master's degree focused on governance and compliance (security policy, ISO 27001, GDPR, law, risk management). Technical track in support: audit, network, pentest. Thesis: deepfake detection with CNNs (XceptionNet).
Amundi · 8 months · Paris
Python automation of Forcepoint (DLP) alerts for the SOC, AI project in collaboration with the Crédit Agricole Data Lab.
TryHackMe · RootMe
Continuous practice on public platforms: web exploitation, forensics, Active Directory, Linux/Windows post-exploitation. Methodology captured into reusable checklists.
Stack
Background
Built production React/TypeScript features on a large-account project. LLM exploration, Agile team (code reviews, Git, sprints).
Python automation of Forcepoint (DLP) alerts for the SOC. AI project in collaboration with the Crédit Agricole Data Lab.
YOLO model for aquatic waste detection (TensorFlow, 91% precision) - pipeline deployed in production.
Contact
Looking for a full-time AI or data engineering role - open to relocation across France and internationally (V.I.E). Also happy to chat about an ML, backend or security project. Quick to reply, in French or English.