AI & LLMs in production
Multi-model agents, RAG, semantic search, guard-railed generation. The step from notebook to a service thousands of people call: latency, cost per request, safeguards, evaluation.
Available for work
I ship AI to
production.
Full-stack and AI engineer, ten years building platforms that hold up under load and AI systems that make it out of the prototype. Agents, RAG, semantic search, optimisation — from the data model to the infrastructure that serves it.
What I do
Four kinds of work, one standard: it has to run for real.
Full-stack platforms
A clean API, a front end that doesn't lie about the state of the system, a database thought through before the first line of code. From design to deployment, alone or leading a team.
Optimisation & algorithms
Assignment, scheduling, forecasting. The problems where a well-framed heuristic beats one more model — and where the gain is measured in euros, not in metrics.
Data & automation
Pipelines, internal ERPs, dashboards that actually inform a decision. Automating what repeats, to give time back to the people doing the creating.
Experience
Ten years of systems brought into the world.
-
2024 → present
AI Engineer — Sanofi
Freelance engagement · Healthcare / Pharma
Designed and shipped, end to end, an agentic AI platform that automates the production of regulated marketing content — emails, articles, manuscripts. Multi-model agents, generation held to the sector's compliance constraints, real-time streaming, AWS infrastructure.
-
2018 → 2024
Lead developer & Project Manager — VINCI
Freelance engagement · Construction · Paris · 6 years
A SaaS platform for managing professional competency frameworks, built and carried end to end: REST API in Go, Angular SPA, storage and processing on AWS (S3, Lambda, SQS, MySQL). Automatic generation of frameworks from business documents, backed by GPT-4o and Claude.
-
2017 → present
Founder — DataHeroes
Consultancy and software studio
The company behind my engagements, and the idea that came with it: giving freelance work a sense of the collective back. Today also a workshop — business applications, data tooling, products of my own.
-
2015 → 2017
CTO — Popchef
Foodtech · Paris
Led the technical team of a fast-growing corporate canteen: hiring and managing the team, the backend, and above all the algorithms that automated daily operations — order forecasting, delivery assignment, an internal ERP designed and grown in place. One meal delivered every five seconds at peak.
-
Earlier
Development & data science
Education: HEC Paris
The first years in the trade, on the development and data science side — the ones where you learn that production constraints always arrive earlier than you expect.
Stack
Tools in service of the problem.
AI
- Multi-model agents
- RAG
- Semantic search
- GPT-4o
- Claude
- Embeddings
- Evaluation
Backend
- Go
- Node.js
- Python
- REST APIs
- MySQL
- PostgreSQL
- Queues
Front end
- Angular
- React
- TypeScript
- SPA
- Streaming UI
Infrastructure
- AWS S3
- Lambda
- SQS
- CloudFront
- CI/CD
- IaC
Data
- Pipelines
- Forecasting
- Optimisation
- Data visualisation
Leading
- Tech lead
- Scoping
- Project management
- Hiring
Contact
An AI project to get out of the prototype?
Tell me the problem before the solution — I'll come back with a view of the path, the cost, and what is going to be hard.