Software engineer · Paris

Systems you
can talk to.

I build AI-integrated products: voice agents, agentic pipelines and the backends behind them. Trained at 42, with a law degree from Panthéon-Assas; I care about how ideas turn into systems that actually run in production.

Selected work

01

Ardoise

Agentic accounting dictation · latest

Dictate, photograph or type an expense: a pipeline of four agents extracts it, codes it (French chart of accounts, via retrieval), drafts the accounting entry and verifies it independently, with a correction loop and human hand-off. OCR of a receipt is just one more input into the same cycle.

  • React
  • TypeScript
  • Python
  • LangGraph
  • RAG
  • EU LLM
02

RAG Assistant

Retrieval over French public services · in Rust

A retrieval-augmented assistant in Rust: hybrid search (dense + BM25, fused with RRF) over pgvector, query expansion, grounded answers with citations. Eval-driven: retrieval hit-rate 73%→100%, MRR 0.67→0.97.

  • Rust
  • axum
  • pgvector
  • ONNX embeddings
  • Docker
View repository
03

Voice Weather Agent

Browser voice agent with tool calls · deployed

A browser voice agent: speech-to-text, LLM reasoning, text-to-speech over a WebSocket, performing tool calls on a custom MCP server wrapping public weather data, with a 3D map interface and a Dockerised backend on Render.

  • Python
  • TypeScript
  • WebSocket
  • MCP
  • Docker

Writing

About

I have worked across software, FinOps and law, which shaped the way I approach technical problems: with structure, curiosity and an analytical mindset. I like building useful systems, understanding the constraints behind them, and I am open to challenging technical environments where I can keep learning fast.

SoftwareBackend, APIs, data systems and deployed products
FinOpsCost visibility, reporting and practical cloud analysis
Analytical path42, law, product thinking and technical challenges