videos
Recorded talks and tutorials by Naman Goyal on long-horizon agents, post-training and reinforcement learning, evaluation, and alignment of large language models, plus teaching material on the foundations.
Agents and Alignment YouTube channel. New talks and tutorials land here first. Subscribe Tutorial A 60-minute blitz through gradients and backprop Multivariate and matrix calculus, the chain rule, and backprop worked by hand. Plays only my part, 6:33 to 1:19:32. Talk Humans + AI: Collaborative Intelligence for Complex Decision-Making When models should defer to humans, how to communicate confidence, and training that makes models better collaborators. A 90-minute workshop. Talk Agentic LLMs in Practice: Tools, Function Calling, and Workflow Design How tool use and function calling work under the hood, graph-based agent workflows, and the production anti-patterns that break them. Talk Taming Non-Determinism: A Framework for Evaluation and Observability in Autonomous Agent Trajectories Evaluating and debugging non-deterministic agent runs: trajectory evals with a judge model, cost and latency trade-offs, and sandboxed execution. Talk The Ascendancy and Challenges of Agentic Large Language Models How LLMs went from text generators to goal-directed agents: planning, memory, tool use, ReAct versus plan-and-execute, and graduated autonomy. Talk Beyond Text Generation: The Rise of Agentic AI and Its Transformative Potential A short tour of agentic AI: technical foundations, multi-agent systems, embodied agency, continual learning, and the ethical questions.