I’m interested in anything that makes training and inference faster, from low-level kernel programming to high-level model architecture design. I also like open source! I’ve made (small) contributions to Huggingface and Triton in the past.