AI & Systems engineer
Computer science student at Télécom SudParis, with a background in data science, AI, networks and systems. I work in Python and Rust: training models, running campus networks, writing Android system daemons.
An open-source continuity layer between Android and Linux, in the spirit of Apple's Continuity. A native Rust daemon inside the ROM handles crypto, QR pairing and routing, then opens an Ed25519 mTLS QUIC tunnel to the desktop.
A deliberately small voice-notes app: one screen to record, one to search. Recordings never leave the phone; speech-to-text runs locally every night, so the archive gets better while you sleep.
A comparative study for my prépa TIPE: Pac-Man learns to play, trained with neuroevolution, deep Q-learning and PPO on a custom game engine, with replays and training curves to compare them.
Built internal tools in PHP and AngularJS that take the repetitive, tedious parts out of employees' daily work, so they can spend their time on what they do best.
Built an online booking space: a REST API in Python with FastAPI and its database, plus a Vue.js interface on top. It served 250+ users, 370+ bookings and 840+ recorded volunteer hours.
Extended a Python program that formats Engineering Change Orders (ECOs), helped build a VBA budget tool for several company entities, and made Power Query reports that track ECO progress, spot bottlenecks and flag late submissions.
Wrote a Python program to automate the processing of Engineering Change Orders (ECOs) and prepared large volumes of data for migration into a new PLM system.
Taught computer science (NSI) to a home-schooled first-year student, adapting the programme to how she learned.
MiNET is the student-run internet provider on campus, serving around 800 residents. As Vice-President for a one-year term, my job was to keep it running and, above all, to teach the next team to run it.
My own Swarm-like orchestrator for Docker. Like Swarm, it runs on Raft consensus; unlike Swarm, it comes with replication, DNS-based high availability, a reverse proxy, a web interface and backups built in. All of it ships as a single 61 MB binary.
Term 1 · Click a node to stop it: the others elect a leader if needed and take over its containers.
Tech is still built by too few kinds of people. I give a real share of my time to showing girls and young women that engineering is theirs too, and to making student life safe for everyone.
Teaching is what I enjoy most. I coached a team of five middle-school girls and taught them, step by step, to design and build their own mobile app, from the first idea to a working prototype. I also helped run the programme's opening and closing ceremonies.
Passing on what I was given matters to me. I went into middle and high schools to share my path into science and engineering and the curiosity behind it, with a clear aim: encouraging girls to see those paths as theirs. I also took part in conferences on gender equality in scientific and technical careers.
For these two clubs at my school, I was in charge of safety at student parties, preventing sexist and sexual violence, and I built SafeINT, the app that supports that work.
The campus events calendar I led at MiNET: students follow associations and clubs, see every event in one place and export their own calendar as ICS. The student union runs all campus events on it.
The safety app for our student parties: ask a volunteer to walk you home, request a call or find a safe place, with live location sharing and walking routes computed over OpenStreetMap.
Draw a multi-node WireGuard network and get keys, configs, nftables and sysctl ready to deploy, all in the browser.
Turn YouTube and YouTube Music into correctly tagged, Jellyfin-ready libraries, with auth delegated to Jellyfin.
AI can draft, but it cannot be accountable. Every line that goes out under my name, I have read, understood and tested, and its bugs are mine to fix.
Some of my projects were built with AI and some were not, recent ones included: marcher is written entirely by hand. Hover any project to see how much and how AI was used.
I use it where it saves time on work I already understand, and leave it out where the point is to learn how something works.
I want to find technical solutions that genuinely help the people who can't build them themselves.
