Learn directly from Will Brown & Kyle Corbitt

Incredible value! Both Will and Kyle did a great job in sharing their practical knowledge!

Great course. I really enjoyed this course! Will and Kyle are great at explaining things in a simple way. What I loved most was how they didn't just teach us the technical stuff - they helped us in getting the right perspective , on how to think about solving problems with Agents and RL. And I really loved the live coding sessions.

learned a ton in this course!

I went into the course hoping for a few tips on reinforcement learning loop, and came out with a deploy-ready playbook: - Hands-on, not hand-wavy. Will and Kyle walked us through live rollouts, reward curves, and hyperparameter sweeps. No fluff—just real code and real patterns that scale. - Tiny-judge trick = instant feedback. Training a lightweight classifier on “edited vs. untouched” outputs turns everyday user edits into a live reward signal. No massive labeling effort needed. - Scale that fits your wallet. From single-GPU LoRA with Unsloth to full-on Ray/VLLM clusters, the infra patterns let you prototype on a 4090 and scale to H100s without rewriting anything. - Tooling that respects engineers. The trainable-model + run-config repo structure, Skypilot launch scripts, and W&B dashboards make it easy to run fast, track results, and iterate with confidence. By the end of Lesson 6, I had a checklist ready to upgrade my current project: wire up quote-edit logging, kick off nightly RL sweeps on my checkout agent, and wrap VLLM behind a gateway to keep GPUs fully utilized. If you’re building real-world agents and want to move fast without cutting corners, this course delivers.

The course delves deeply into constructing agents without relying on abstractions or frameworks, focusing on foundational patterns and core principles.