Week 1 (Friday): Foundations - the reliable reasoning component
8 lessons · Back to full syllabus
What you keep
How to turn a model call into a dependable, observable piece of software.
You ship
A working `/ask` endpoint that answers reliably - the reasoning core of your capstone.
Watch through (async)
Think Like an AI Engineer (free 10-part series)
Watch this YouTube playlist before or alongside Week 1. Plain-English foundations - how LLMs work, prompting, models, context, agents, workflows, and evals - so the live FastAPI and reliability work clicks faster.
- 1How LLM Actually Works (in Plain English)Watch
- 2How to Prompt AI Like a Pro: The 5-Part FormulaWatch
- 3How to Pick the Right AI Model for the JobWatch
- 4Using AI With Your Files, Images and DataWatch
- 5Context Is Everything: The #1 Skill for Using AIWatch
- 6Vibe Coding for Non-Engineers (and Engineers)Watch
- 7What AI Agents Actually Are (Hype vs Reality)Watch
- 8Build AI Workflows With No Code (n8n, Zapier, Make)Watch
- 9AI Evals: How to Know If Your AI Actually WorksWatch
- 10The Fastest Way to Stay Current in AIWatch
Live session resources
Lessons
From model call to reliable component
Production means the right shape, at a known cost, with a plan for when it fails - every time.
The OpenAI API and the Playground
Experiment in the Playground first: system vs user messages, temperature, max tokens, and the token counter.
FastAPI and the /ask endpoint
Wrap the model in typed software: POST /ask returns the answer plus tokens used.
Prompting with rigor
Structured outputs, few-shot, and decomposition - tested live against your own endpoint.
Output guardrails and validation
Validate the response matches your schema; reject or retry on malformed output.
Token economics
Treat the context window as a budget - few-shot, system prompts, and verbose outputs all cost on every call.
Model selection
Match capability, cost, and latency - swap models on the same endpoint and watch tradeoffs move.
Ship your reliable reasoning endpoint
Schema-validated answer, token usage, guardrail on malformed output, and a justified model choice.
Lessons in this module