Stas Bekman
Software engineer focused on LLM training at Snowflake AI Research; author of an open ML engineering book
- Develops efficient post-training software at Snowflake AI Research, Snowflake AI Research12
Stas Bekman is a software engineer who specializes in training large language and multimodal models in the PyTorch ecosystem and currently works at Snowflake AI Research on efficient post-training software. Bekman's site describes earlier work on a distributed training framework at Contextual AI and, before that, time at Hugging Face from 2020 to 2023, where Bekman was lead engineer on training the BLOOM model and later helped train IDEFICS-80b. Bekman is first author of Arctic Long Sequence Training (2025), which the paper says is open-sourced via DeepSpeed, and maintains the Machine Learning Engineering Open Book under a CC BY-SA 4.0 license.12346
Work
Arctic Long Sequence Training (ALST)Open-source code
First author of the 2025 paper on ALST, a set of memory optimizations for training on very long sequences with Hugging Face models; the paper says ALST is open-sourced via DeepSpeed and Snowflake's Arctic Training.3
Machine Learning Engineering Open Book (external site: github.com)Published research
Author and maintainer of an open collection of methods, tools and step-by-step instructions for training, fine-tuning and serving large language and multimodal models, distributed under CC BY-SA 4.0.41
Universal Checkpointing (external site: arxiv.org)Published research
Co-author of the 2024 paper on Universal Checkpointing, a distributed checkpointing system for large-scale training with reconfigurable parallelism.5
OBELICS and IDEFICS (external site: huggingface.co)Open data
Equal-contribution author, with a Hugging Face affiliation, of the 2023 OBELICS paper, which releases an open interleaved image-text dataset and trains the IDEFICS models.6
Other documented roles
Roles the sources document but that are not confirmed as current.
- Worked on a distributed training framework and on training and inference of RAG LLMs, Contextual AI1
- Worked at Hugging Face (2020–2023); lead engineer on the BLOOM training, Hugging Face16
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.