Advanced Prompting and Reasoning
For engineers who already prompt LLMs and now need multi-step reasoning they can trust in production. You will build tool-using agent loops, use sampling, search, debate and critique where they pay for themselves, and evaluate reasoning step by step rather than by its final answer alone.
Course Content
Ground a model's reasoning in real tool results, raise its reliability by voting over independent samples, and search over partial solutions when early choices constrain later ones.
Make critique, debate, tool use and reasoning evaluation reliable by bringing in outside evidence and measuring what each step actually changes.
Build a research agent that plans before it searches, gathers quoted evidence under a budget, verifies every claim in code, and says plainly what it could not find.