Building @UnslothAI. Finetune LLMs 30x faster http://github.com/unslothai/unsloth. Prev ML at NVIDIA. Hyperlearn used by NASA. I like maths, making code go fast