Strong Compute removes bottlenecks in ML training pipelines by applying low-level CPU/GPU optimizations and tooling that give a detailed view of how users' code interacts with hardware. The company says its optimizations can speed up training by 10x to 1,000x depending on model, pipeline and framework, and recently made an Nvidia reference implementation run 20x faster for customer LayerJot. Founded by Ben Sand and part of Y Combinator’s Winter ’22 batch, Strong Compute currently has six full-time engineers and plans to double headcount over the next few months. The team is hiring mathematicians and building tools to automate many current manual optimization tasks. Its immediate commercial focus remains on computer vision while pursuing development partners in self-driving, medical and aerial sectors. The company notes inbound interest from large customers that sometimes spend $50M or more on compute and describes the market as bimodal (customers spending either <$1M or $10M–$100M). It recently raised funding to support longer-term R&D rather than short sprint deliverables.

Total raised

$8M

Funding rounds

1

Latest round

Seed

Latest activity

May 2022