Building AI Agents

What an AI agent actually is, and how to build one: a language model, a set of tools, a loop and some memory. Starting from the wire format a model speaks, you will write the agent loop by hand before reaching for any framework — tool definitions and dispatch, the Model Context Protocol, retrieval, evaluation. By the end you will know which problems deserve an agent, and which are better served by a single call.

Everything here is open. Sign in and enter your teacher’s group code if you want your homework counted.

• Lecture 1. From LLM to AI Agents2026-10-08

  In class: py_functional/warm_up, py_functional/profiler, Practice 1 — How to use LLMs (Colab), Practice 2 — Your first agent (Colab)

• Lecture 2. Tools and MCP2026-10-15

  In class: py_functional/lru_cache, py_functional/context_manager, Practice — Tools and MCP (Colab)

• Lecture 3. Memory and Guardrails2026-10-22

  In class: agents_retrieval/bm25_index, agents_training/lora_adapter, agents_retrieval/email_regex, Practice — Memory and Guardrails (Colab)

• Lecture 4. Multimodality2026-10-29

  In class: py_classes/range, Practice — Multi-Agent Systems and Multimodality (Colab)

• Lecture 5. Agent Evaluation: From Metrics to Managed Quality2026-11-05