Software engineer
Extend existing software with autonomous workflows.
Stanford / Stanford Online
Build AI systemsAgentic AI
Build agent systems through programming assignments adapted from Stanford graduate teaching.
Curriculum
- Test-time scaling and self-improvement; tool use, reasoning and planning
- Retrieval and long-term memory
- Evaluation and long-horizon tasks
Before you start
Independent Python programming; LLM API experience; foundational ML knowledge. Short fit application.
Who it suits
Our assessment, based on the syllabus.
AI or ML engineer
Build and evaluate agents beyond basic prompting.
CTO or engineering VP
Assess architecture and reliability through implementation.
Expected outcomes
- Build a small agent through coding assignments.
- Evaluate agent behaviour and extend self-improving workflows.
- Assess the roles of tools, memory and planning.
Example application
Build a prototype that calls a tool, then check how it handles failed or incomplete results.
Limits
- Python/API prerequisites exclude nontechnical beginners.
- Approximately 30 hours cannot establish full production expertise.
How current is it?
Enrollment opened September 2026 per Stanford’s program landing page; current syllabus includes modern agent research.
When to choose it
Consider this if you already use Python and APIs and need to build agents. The programming prerequisites make it unsuitable as a first AI course.
Sources
- Official course page ↗online.stanford.edu
- Official supporting source ↗learn.stanford.edu
Checked 20 October 2026. Confirm the current offer on the provider’s page before paying.