capecape

I’m Thomas Capelle (cape), Staff Machine Learning Engineer at Weights & Biases (now part of CoreWeave), on the AI Applied Team.

These days I build Senpai, an autonomous ML research framework (ICML 2026), and use it on CFD surrogate modeling (AirfRANS, TandemFoilSet, DrivAerML) and aerodynamics with the Aston Martin Aramco Formula One Team. Most of the interesting work is flow matching and diffusion models for pressure and velocity fields.

Before that I spent several years on LLMs: fine-tuning, aligning, evaluating, quantizing, deploying, and making wandb useful to practitioners through wandb/examples, MLOps content and Fully Connected. Earlier still I used deep learning for short-term solar forecasting at Steady Sun. Background in Urban Planning, Combinatorial Optimization, Transportation Economics and Applied Math.

Profile photo

Experience

Weights & Biases by CoreWeave

Staff ML Engineer (2021–present)

  • Senpai — an autonomous ML research framework: advisor and student agents on Kubernetes coordinating entirely through GitHub PRs and labels, with W&B as the experiment record (ICML poster)
  • State-of-the-art CFD surrogate modeling with the Aston Martin Aramco Formula One Team — neural PDE surrogates for pressure and velocity fields, with flow matching / diffusion residual models to recover high-frequency detail and quantify uncertainty
  • Large-scale distributed training on GPU clusters: elastic multi-node runs, fault tolerance, checkpointing and resilient recovery
  • LLMs everywhere: fine-tuning, aligning, using, deploying, quantizing, eating for breakfast
  • Improving user experience through examples in wandb/examples
  • Creating content for Fully Connected on how to make machine learning better
  • Helping companies integrate wandb into their MLOps workflows

Steady Sun

ML Engineer (2020-2021)

  • Developing next version of short term forecasting algorithm based on sky imager
  • Integration of ML tools into the Steady Sun code base
  • R&D in deep learning for detection/segmentation and forecasting of sky images
  • Collaboration with NVIDIA AI on deployment and production of DL models

INES CEA

Research Engineer (2018-2020)

  • Development of algorithms for failure detection in photovoltaic systems
  • Deep learning model for parameter regression of IV-curves of photovoltaic modules
  • Thermal image segmentation and classification model
  • Project engineer and coordinator for a Franco-Chilean partnership, international cooperation with multidisciplinary laboratories

Inria

Research Engineer (2017-2018)

  • Integration of spatial models with air quality and forecast models

Inria STEEP

PhD Researcher (2013-2016)

  • Development of mathematical frameworks for urban/transport model calibration
  • Multidisciplinary team work (engineers, economists, urbanists)
  • Supervision of interns
  • International conferences and peer-reviewed publications

CIRRELT

Master Intern @ Polytech Montréal (2011)

Universidad de los Andes

Assistant Professor (2010)

  • Analysis and algebra for engineers

Publications

The full list of my publications is available via Google Scholar

Education

PhD in Computer Science
Université de Grenoble
2017

MSc in Civil Engineering - Transport
Universidad de Chile
2013

Mathematical Engineer
Universidad de Chile
2013

Languages

  • Spanish (mother tongue)
  • French (native)
  • English (full professional proficiency C1)

Download LaTeX CV | Kaggle Profile