Optimize, Verify, Build Trust

Writing AI GPU inference kernels, optimizing for performance, and catching the ones that lie. AI is here to stay, gains more power by the day, alignment in safety and accuracy is paramount.

Short posts on the ideas underneath the series, each with a hands-on exercise: build it by hand, then check it against a real library. Self-paced, so they go up as they're finished. Each links its code in the llm-foundations repo.