Just started fall quarter. The weather has been pretty pleasant in general!

Making an inference engine! I'm trying to learn about how model inference works, and I think making my own engine is the best way to do so :). I think I'm already somewhat competitive with vLLM / SGLang on the model I'm starting with (Qwen3 1.7B on an RTX 3090), but I want to do less "vanilla" models and make them fast.