Charlotte Caucheteux

 

I am a Senior Research Scientist at Google DeepMind, where I lead the technical direction of the AI for Health team in Paris. Driven to create tangible impact in medicine, our team builds models and agents for medical applications, with a primary focus on oncology. Prior to this, I contributed to Gemini's pre-training to strengthen its foundational medical capabilities.

Between 2023 and 2025, I was a Research Scientist at Meta AI (GenAI), where I led the Memory effort focused on LLaMA long-context and retrieval capabilities, and developed zero-shot tool use for LLaMA 3.

I hold a PhD in deep learning from Meta AI (FAIR) and Inria (investigating neural vs. artificial language representations) and graduated from École Polytechnique. Earlier in my career, I worked in healthcare AI at Owkin and AP-HP, and co-founded the AI-art collective NoArtist.

My research interests include Large Language Models, Neuroscience, and AI for healthcare & biology.

Feel free to connect on LinkedIn or Twitter.

profile photo
Selected Publications

I am interested in the computational basis of natural and artificial intelligence, with a focus on language. As part of my PhD, I investigated language representations in deep neural networks and the human brain, using transformer-based language models and neuroimaging techniques (fMRI, MEG, and EEG). My publication history can be found on Google Scholar.

Preprints & Journal Articles

Conference Articles

Press Coverage

Selected Threads

Misc.

  • I received an MSc/BSc from Ecole Polytechnique (diplôme d'Ingénieur), with a specialisation in Machine Learning.

  • I have a strong interest in Healthcare. Prior to my PhD, I had the chance to work for Owkin, a start-up (now unicorn!) applying AI to oncology (2017). I also worked for the Assistance Publique–Hôpitaux de Paris (AP-HP) during the COVID crisis (2019).

  • Finally, I enjoy arts 🎨 and music 🎹. In 2019, two teammates and I founded NoArtist, a collective of AI-generated artworks. We regularly organise exhibitions and events to democratise AI through arts.