The program
What students leave with.
The skills every student can show by Demo Day, and the tools they build with.
What students can do by Demo Day.
Pick the right model
What models cost, how they fail, how to compare them
Typed output and tool calls
Reliable structure, real APIs, a human approves risky actions
RAG with citations
Answers grounded in a company's own documents, and an honest "I don't know"
Evals, not guesswork
Measure quality in CI, improve with evidence
Agents that act
Multi-step tools, outside systems through MCP, and when a workflow is better
Run it in production
Docker, logs, cost and latency
Ship it safely
Guardrails, PII, secrets, prompt injection, a human in the loop
Scope and ship like a Forward Deployed Engineer
Discovery, a design doc, a deadline, working software handed over
Maps to NACE's Technology, Critical thinking and Communication competencies. The weeks 3–6 coaching sessions cover the other five.
What “functioning” and “explain” mean.
Each day adds one capability to the same app, so the capstone is done by day 10.
The stack students build with.
Each student's app is one Python project, built from an empty folder and live on Vercel on day 1, in a project the program sets up with the student's model key already in it. On day 7 students package it with Docker, so in week 5 it can run in the partner's own cloud. The AI Gateway gives one key for several model providers, so students compare models instead of learning one vendor. The real costs are model usage and, depending on the program's plan, Vercel team seats.