Experience

  1. Independent AI/ML consultant

    August 2025 to present

    Remote: United States to February 2026, France since. RAG and agent architecture, LLM pipelines, evaluation, production forecasting. Expert-Vetted on Upwork, five stars. Availability: two to three days a week of client work.

  2. Co-founder, GENAU AI

    November 2024 to present

    I led the technical build and coordinated a five-person team, standing up the production platform from scratch: OVH compute, Railway-hosted database, Prefect, Argo CD and GitOps delivery, GCP, VPN, multi-tenant Streamlit dashboards. I also built an in-house team of AI coding agents that turns one command into a deployed pipeline, with verified hand-offs and human approval gates. The models outperform incumbent solutions by 30% on standard benchmarks.

  3. Power Market ML Engineer & Quantitative Analyst — BCM Energy (Planète Oui, now Elmy)

    September 2019 – March 2024

    Lyon, within the trading team. A CIFRE arrangement agreed at signature: one role, whose research strand was the doctorate.

    • Day-ahead price forecasting on European power markets. Improved predictive accuracy by 46%, and built the model-driven trading strategies the desk used daily.
    • I ran my own trading book, tied to the performance of my own model. Very few machine-learning people carry that consequence.
    • Integrated more than 100 variables from dozens of external sources into the internal database, ENTSO-E included.
    • Put the models into production and kept them there: Airflow, Docker, PostgreSQL, Jenkins, GitHub CI, Grafana.
    • Reverse-engineered the market-clearing algorithm and embedded it inside the model — the research is described on the about page.
  4. Research intern — BCM Energy

    March – August 2019

    Lyon. A state of the art on electricity price forecasting, and an end-to-end system from live ingestion to published results. It became the thesis.

  5. University teaching

    2020 to 2023

    Eighty-two hours across three French engineering schools: the EUPHEMIA market-clearing algorithm at M2 level at École Centrale, four years running (32 h); algorithmics and Python at IUT Lyon 1 (42 h) and INSA Lyon (8 h).