About Felafax
Felafax is building open-source software for training and fine-tuning large language models on chips other than Nvidia GPUs, including Google TPUs, AWS Trainium, and AMD and Intel GPUs. Most machine learning tooling is written for Nvidia hardware, so running on alternatives usually takes extra engineering, which Felafax aims to remove. Its public work includes a fine-tuning of the 405 billion parameter Llama 3.1 model on an AMD MI300X server using JAX. The company was started by twin brothers Nikhil and Nithin Sonti, who earlier spent about five years at Google and Meta on AI infrastructure, and it joined Y Combinator in the Summer 2024 batch.
Felafax was founded in 2024 and is based in San Francisco, CA, United States. It is listed under AI Accelerators. Reported funding is YC S24, $500K seed.
Compute Technologies
- AI Accelerators
What Felafax works on
AI Accelerators
An AI accelerator is a chip, or a part of a chip, designed to run machine learning faster or with less energy than a general-purpose processor. The category includes GPUs, TPUs, neural processing units (NPUs) and chips from many startups.
Also in United States
Oxmiq Labs, Lemurian Labs, Crystal Group, Tesla Dojo, deepsilicon
