Julien Chaumond

Cofounder and CTO, Hugging Face

Search every verified Julien Chaumond interview, podcast appearance, and on-the-record quote — each transcript cross-checked by AI and human review to confirm speaker identity. Julien Chaumond, cofounder and CTO of Hugging Face, has described the company’s origin as a joke about being the first to go public with an emoji instead of a ticker symbol, noting that the hugging face emoji was chosen as a favorite and initially intended to be temporary before community adoption led the company to keep it. He has recounted that the first version of the Transformers library was a PyTorch implementation of BERT, and that the team quickly built a system for users to upload model checkpoints using pre-signed S3 URLs, which was “super messy” but validated interest in model sharing. Chaumond has emphasized the shift in machine learning from training models from scratch to using pre-trained models fine-tuned for specific tasks, calling transfer learning a “much more efficient way of doing machine learning.” He has stated that Hugging Face built its platform around this change, using Git as the backbone for versioning models and datasets to avoid vendor lock-in, and that the company’s goal is to bridge the gap between research and production by providing tooling for sharing, discovering, training, and deploying models.