Building LLMs from scratch (tokenizer, transformer, pretraining, fine-tuning) and applying them to real product management problems, starting with resume feedback. Also interested in running local LLMs, small language models, and AI-assisted product analytics.