From learning by doing, to PhD, to MLE

1. MLE (2023 - today)

I’m a Member of Technical Staff, Machine Learning Engineer (MLE) at Cohere (from June 2025). I’m part of the Applied Machine Learning team, which adapts foundation models to specific languages or tasks, and integrates them into production. Before joining Cohere, I was an MLE at Hugging Face (Dec. 2023 - June 2025), where I helped customers build solutions to real-world problems with ML in a wide range of industries (finance, health, manufacturing, visual media, retail, public sector) for different use-cases (RAG, complex LLM/VLM systems, text/image search, text/image classification) along the full ML life cycle (use-case definition, data annotation, fine-tuning, evaluation, deployment, inference optimization).

2. PhD (2021 - 2023)

I did a PhD on making text classifiers work better in real-world settings. My research was at the intersection between NLP and computational social sciences, with a focus on using transfer learning to make text classifiers produce more valid and more robust outputs with less training data. The models I developed have been downloaded +100 million times via the Hugging Face hub (between December 2021 - February 2025) and two of my models were ranked by IBM in the top 10 of models on the HF hub for downstream classification fine-tuning. See the final thesis PDF here. I also worked as a freelance data science / NLP consultant for different software companies during my PhD.

3. Data Science Consulting (2018 - 2021)

Before my PhD, I worked as a data scientist and researcher at the Centre for European Policy Studies (CEPS) and as a technology consultant for CARSA. For example, I analysed free-text data from citizen consultations for the European Commission using SBERT models, or I co-created the CEPS Eurlex dataset of 142 000+ legal texts with metadata. I co-founded the European Policy Data Science Network for sharing best data science practices in the non-profit sector. I also contributed to several studies on technology policy, such as the official impact assessment of the EU AI Act.

In 2018, I co-founded the NGO ‘DataMine Europe’ (initially ‘European Elections Stats’) as a free-time project to create an automatic seat projection for the European Parliament elections and to improve my programming skills. We produced data and visualizations that were used by leading French, German, Spanish, U.S. media.

4. Studies (BA + MA)

I did a BA and a MA in different social sciences (political science, sociology, economics, law, quantitative and qualitative methods). I initially focussed on empirical social science methods and technology policy, and then dove deeper into the technologies themselves. I learned programming mostly auto-didactically, first in R and later in Python. I owe a lot to the contributors of library documentation, educational blogs and online fora.

Details

You can find more details in my PDF CV