Bio
Team and tech lead for pre-training at poolside, where he trains large language models for code. Recently led the pre-training of Laguna XS.2 and M.1, poolside's first two public open-weight models. Before that, Robert worked as a Senior Researcher at AssemblyAI where he trained multilingual speech models, and previously built AI for cancer and infectious-disease research at InstaDeep and BioNTech's joint lab. MSc in Machine Learning from UCL.
Session (1)
Additional Speaker Details
Using publicly available information we constructed an analysis to help you get a feel for this speaker before deciding to attend their session.
Robert McHardy is team and tech lead for pre-training at poolside, where he sets technical direction for architecture research and distributed training and is one of the authors behind the Laguna family of mixture-of-experts coding models — trained from scratch on 30+ trillion tokens across 6,144 NVIDIA Hopper GPUs. His perspective is grounded in the hands-on work of building and running foundation-model pre-training at production scale.
Speaking style
GitHub
@RobMcHRecent writing (1)
- Laguna M.1/XS.2 Technical Report · paper · May 2026