2026-AI-Development

Winter 2026 AI Development Training Series

This course was developed in partnership between the University of Chicago's Career Advancement Office and the University of Chicago's Data Science Institute.

Workshop: AI Development (4-part series)

This repository contains materials for a four-part workshop on AI development for advanced undergraduates (3rd/4th year).

Lecture 1 (slides-first)

• Topic: Foundations

• Slides: lecture_1/slides/lecture_1.pdf

• Readings for lecture 2:

  • Context Engineering

    • Effective Context Engineering for AI Agents - Anthropic's guide to designing context windows for reliable agent behavior.
    • Context Engineering - Simon Willison's take.
    • Agent Best Practices - Cursor's guide to context engineering.

  • Welcome to Gas Town - LONG, only read as long as you are interested!

  • Gary Marcus on AI Limitations

    • Why ChatGPT Can't Be Trusted With - Discussion of reliability issues and when not to trust LLM outputs.
    • Breaking: Marcus Weighs In, Mostly - Critical analysis of AI capabilities and limitations.
    • Let's Be Honest: Generative AI Isn't - Perspective on what generative AI can and cannot do reliably.

Under Construction below!

Lecture 2 (HITL + evaluation)

• Problem: Marketing automation / outreach drafting with compliance
• Notebook: lecture_2/notebooks/lecture_2_marketing_hitl.ipynb
• Data/docs: lecture_2/data/ (leads + brand + product one-pager + rubric)

Lecture 3 (grounding + API actions)

• Problem: Support ticket triage with grounded replies and safe external actions
• Notebook: lecture_3/notebooks/lecture_3_support_triage_api.ipynb
• Data/docs: lecture_3/data/ (tickets + KB + stub API data)

Lecture 4 (tools/MCP-like + orchestrator)

• Problem: Research brief generation using a tool registry and a simple orchestrator loop
• Notebook: lecture_4/notebooks/lecture_4_tooling_orchestrator.ipynb
• Data/docs: lecture_4/data/ (docs + market signals + tool registry)

Running notebooks (per lecture)

Each lecture directory contains a Makefile, Dockerfile, and pyproject.toml.

From a lecture directory (e.g. lecture_2/):

• make build
• make notebook (starts Jupyter in Docker on port 8888)
• make interactive (drops you into a bash shell in the container)

Building slides

Slides are in Beamer (LaTeX) under each lecture’s slides/ folder.

Example:

cd lecture_1/slides
pdflatex lecture_1.tex