June 2020 – Present

Freelance data scientist & ML/AI engineer

I worked on data science applied to environmental sciences. I was in charge of machine learning at Jane’s Weather.

I developed a digital service for accounting audits. It analyzes huge transaction volumes to detect errors and provides a collaborative web interface with financial statements, indicators, charts, conclusion drafting, …

September 2007 – June 2020

Research scientist in applied mathematics | INRIA Paris

I worked on data assimilation (state estimation, inverse modeling), uncertainty quantification, online learning, ensemble forecasting, probabilistic forecasting, meta-modeling (Gaussian processes). My work was applied to high-dimensional complex environmental systems (air quality, wildland fires, meteorology, renewable energies, noise pollution, smart cities, road traffic).

I published about 55 papers in leading scientific journals, and about 25 conference papers. I supervised 9 PhD theses, and many master students, engineers and post-doctoral students.

December 2005 – August 2007

Research scientist in applied mathematics | École Nationale des Ponts et Chaussées

I was working on data assimilation (variational and sequential data assimilation) and ensemble forecasting in the context of air pollution forecasting, based on simulations of chemistry-transport models at continental and smaller scales.

2002 – 2005

PhD in applied mathematics

École Nationale des Ponts et Chaussées

“Uncertainty quantification and ensemble forecast with a chemistry-transport model”, under the supervision of Bruno Sportisse

1999 – 2002

Master of Science & Engineering degree in applied mathematics

École Centrale de Lyon

2001 – 2002

Master of Science in numerical analysis, PDEs and scientific computing

École Normale Supérieure de Lyon

Languages

English (fluent), French (native) & Spanish (conversational)

Computer science

Python, Scipy, Pandas, Matplotlib, Plotly, Scikit-learn, PyTorch, Tensorflow, FastAPI, Flask, PyQt/PySide, SQLAlchemy, Unsloth.
Deployment: Docker, Kubernetes, Argo Workflows, AWS, Langfuse.
Storage: PostgreSQL, MongoDB, S3, R2.
C++, scientific programming, MPI, OpenMP, Qt.
JavaScript, Node.js, React, Material-UI, Redux, D3.js, OpenLayers.
Occasional use of Lua, Lisp, Fortran.
Linux, Ubuntu, Debian, server admin.
Development of large software projects, experience with design and development of multiple libraries with over 50,000 lines of code
Development tools: Emacs, Git, Poetry, GCC, SCons.
Other software: QGis, LaTeX.
Data sources: Hugging Face, OpenStreetMap, ECMWF, NCEP/GFS, Copernicus CAMS, Landsat, Sentinel.

AI Public Portfolio

Word teacher, API generating cloze quizzes for vocabulary learning in English, French, and Spanish.

• Multilingual FastAPI application with LLM-generated quizzes, PostgreSQL storage, pre-generation workflows, and Langfuse tracing
• Reduced quiz ambiguity with masked-language-model ranking and offline evaluation
• LoRA fine-tuning

News curator, extracts novel information from daily news and generates dense daily digests.

• End-to-end LLM pipeline for RSS ingestion, article extraction, takeaway and digest generation
• Embedding- and cross-encoder-based deduplication with PostgreSQL/pgvector

Mathematics & data

Data assimilation and inverse modeling, variational assimilation, reduced Kalman filtering, minimax filtering.
Uncertainty quantification, sensitivity analysis (Sobol’), Monte Carlo, MCMC, ensemble scores, risk assessment.
Forecasting, online learning, sequential aggregation, expert selection or aggregation, filtering, probabilistic forecasts.
Machine learning, deep learning (neural networks), Gaussian processes, RBFs, statistical emulation.
Geostatistics, kriging, covariance modeling, interpolation in high dimension.

Application fields

Environment, air quality, wildland fires, meteorology, renewable energies, noise pollution, smart cities, road traffic
Accounting and audit

Supervision

9 PhD students
5 post-doctoral students, 13 master students
11 engineers

Other achievements

1 startup co-founded: Ambiciti
1 French patent (ensemble forecasting for renewable energies)