Agentic Machine Learning Systems

Syllabus · In preparation

Textbook Volume III: Agentic Machine Learning Systems, a working draft
Prerequisites Applied experience with large language models, and systems programming
Format A draft plan for a 16-week semester or a 10-week quarter
Materials Slides

Description

Students study the systems that run AI agents: models that plan over many steps, keep memory across them, call tools, and recover when a step fails. Volume III covers these systems in twenty chapters, grouped into seven parts.

Reading schedule

These materials are still in preparation for teaching in the next semester or year, so they will change as the course takes shape.

Four pairs of related chapters share a week.

Week Reading Slides
1 Introduction Ch 1
Model Principles
2 The Foundation Model Ch 2
3 Test-Time Compute Ch 3
Agent Memory Principles
4 Context Engineering Ch 4
5 KV Cache Management Ch 5
6 Long-Term Memory Ch 6
Tool Use Principles
7 Tool Calling Ch 7
8 Agent Sandboxes · Agent Security Ch 8 · Ch 9
Agent Runtime Principles
9 The Agent Harness Ch 10
10 Durable Execution · Failure Recovery Ch 11 · Ch 12
11 Agent Evaluation Ch 13
12 Operating Agents Ch 14
Trajectory Learning Principles
13 Trajectory Curation · Trajectory Fine-Tuning Ch 15 · Ch 16
14 Reinforcement Learning from Verifiable Rewards Ch 17
Agents at Scale Principles
15 Multi-Agent Coordination · Agent Economics Ch 18 · Ch 19
Synthesis Principles
16 Conclusion Ch 20

Status

The course has not yet been taught, and it has no labs or offering yet. The schedule is a draft that follows the chapter order of Volume III.

Back to top