Agentic AI Program Overview

Learn more: https://stanford.io/4iHlGQp

Build modern AI agents with Stanford's graduate-level curriculum, adapted for working professionals.

Taught by Stanford Adjunct Professor Chowdhery and Assistant Professor Mirhoseini, this Agentic AI program takes you beyond basic LLM prompting to agents that reason, plan, use tools, and improve themselves through interaction with their environment.

What you'll learn:
- Design agentic systems for complex, multi-step real-world tasks
- Apply test-time scaling to boost LLM performance
- Use self-improvement methods: verifiers, feedback loops, RL, and search
- Build agents that use tools and take actions effectively
- Add retrieval and long-term memory to LLMs
- Develop planning and multi-step reasoning
- Evaluate agent performance with robust frameworks

Drawing on Stanford research and the latest advances in the field.

About the Instructors
Stanford Engineering Adjunct Professor Dr. Aakanksha Chowdhery led end-to-end training of the 540B PaLM model (the largest densely trained language model in the world at the time) and drove pre-training and scaling of Gemini's mixture-of-experts models. Stanford Assistant Professor Azalia Mirhoseini co-developed the MoE architectures now used in nearly every frontier model, created AlphaChip (the RL method behind Google's TPU designs), and pioneered LLM test-time scaling; she worked on both Claude and Gemini.