Thomas Chaigneau
Senior AI Engineer at CHAPTR
France
Summary
🤩
Rockstar🎓
Top SchoolThomas Chaigneau is a founder and machine learning engineer with seven years of experience turning research-grade AI into reliable, production-ready systems. He builds focused infrastructure checks and realistic benchmarks at Kubeply to prevent costly outages, GPU waste, and slow releases—specializing in deployment safety, observability, secrets, and recovery for Kubernetes and GPU workloads. Previously at Owkin he delivered a RAG system that tripled biomedical research velocity and contributed ONNX support for multiple transformer models in the widely used Hugging Face transformers repo. Thomas blends hands-on model training and deployment (from ASR to XXL LLMs) with practical tooling—he’s shipped ML pipelines, optimized multi-GPU training, and authored internal SDKs and benchmarks to reduce toil. Based in France, he uniquely frames infra problems through agent evaluations, teardown notes, and focused automations that turn recurring pain into actionable runbooks and remediation.
6 years of coding experience
5 years of employment as a software developer
Développeur IA, Intelligence artificielle, Développeur IA, Intelligence artificielle at Ecole Microsoft IAMaster MEEF PE, Master MEEF PE at INSPE de BrestLicence de biologie, Sciences et Vie de la Terre et de l'Univers, Licence de biologie, Sciences et Vie de la Terre et de l'Univers at Université de Picardie Jules Verne (Amiens)
English, French
Github Skills (10)
transformers10
pytorch10
machine-learning10
nlp10
onnx10
python9
language-model9
hub8
tensorflow4
jax4
Programming languages (9)
Github contributions (5)
huggingface/transformers
Oct 2021 - Aug 2022
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:
ML Engineer
ML EngineerContributions:45 reviews, 10 commits, 10 PRs in 9 months
Contributions summary:Thomas primarily contributed to the integration of new models and configurations within the `transformers` library, specifically focusing on supporting ONNX export for various models, including Camembert, Flaubert, and ConvBERT. Their work involved modifying configuration files, updating feature mappings, and adding necessary type hints to enable compatibility with ONNX for these transformer models. They also addressed issues related to ONNX conversion for GPT-J, ensuring correct data types and configuration.
pythonbertspeech-recognitionstate-of-the-artflax
Wordcab/wordcab-transcribe
Mar 2023 - Mar 2024
Contributions:7 releases, 50 reviews, 139 PRs in 1 year
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.