CI/CD for GPU sensor fusion
Master's thesis. A GitHub Actions pipeline for CUDA sensor fusion on an NVIDIA AGX Orin.
Software & AI engineer building in Brussels
I'm Moussa. I like building things, from the first sketch to the moment someone actually uses them. Backend, frontend, AI, teaching: I learn by touching a bit of everything. But first, here's some work:
Master's thesis. A GitHub Actions pipeline for CUDA sensor fusion on an NVIDIA AGX Orin.
Scanned language textbooks into structured vocabulary. Mistral OCR with JSON‑schema output.
A second-hand marketplace. I led the architecture and owned the backend.
REST API for campus navigation, presented to the public at Printemps des Sciences.
Networked board game in C++ with SFML and raw sockets. Team of 8.
Large-scale traffic prediction with distributed processing on Spark.
Experience
Designed and built a billing feature end to end: architecture, REST API, database schema, Spring Boot backend and Angular UI. Worked directly with the non-technical founders on requirements and demos. In production, used by 50+ people.
Legal AI assistant for Belgian law: scraping and ingesting legal texts, RAG retrieval, Django REST backend, Vue.js frontend. It taught me a lot about data quality and RAG.
End-to-end test automation for a Spring Boot application with Cypress and Cucumber (BDD), in a Scrum team.
Exercise sessions for first-year Programming Languages: complete beginners, mixed levels, one-on-one help and grading.
What I build with AI
Turns scanned language textbooks into structured vocabulary sheets. Mistral OCR reads each page and returns one JSON entry per vocabulary block, shaped by a JSON‑schema.
Scraping and ingesting legal texts, then retrieving the relevant ones to ground the model's answers. It taught me where RAG really breaks: data quality.
Working with MCP, LLM tool calling and prompt engineering, and using Claude Code every day in my development workflow.
CI/CD for GPU‑accelerated multi‑sensor fusion on an NVIDIA AGX Orin: building, packaging and validating CUDA code automatically.
Intro‑to‑AI workshops for families at public events in Brussels.
Where I come from
I did my bachelor's in Computer Science at ULB, and I'm finishing my master's at VUB (burgerlijk ingenieur track). Backend is where I learned to build. AI is what I can't stop building with now.
It started with RAG at Hayolabs and never really stopped. It's my evenings too: a personal project that turns scanned textbooks into structured vocabulary, new vision models to try as soon as they come out, Claude Code open every day.
I like it enough to explain it on weekends: with Genius Generation I run intro to AI workshops for families at public events in Brussels.
How I work
Talk to the people who'll use it. At Arqaam, that meant working directly with the founders on what billing actually needed.
Architecture, API and schema before code. Boring decisions early save painful ones later.
Whatever the stack, it goes to production, with CI doing the repetitive part.
TA at ULB and workshops for families. Explaining something is the best test that I get it.
Stack
Based in Brussels and available right away. Happy to talk about any engineering role: software, AI, data or infrastructure.