Indie Degree
GitHub
← Programme

Architecture and Judgement

AIE-107

2 credits · 32h required · 1h optional · after AIE-101, AIE-102

Every other course in this block teaches you to build one thing well. None of them ever asks whether you should have built it at all. This one does nothing else. A customer says 'can AI do this' and the valuable answer is frequently 'you do not need AI for this, you need a rule and a lookup table' — and being the person who says that, with numbers, is worth more than being the person who can wire up any architecture on request.

The lab. Career Side Quests has a clean classification problem hiding in it — given a line from a job posting, is it a requirement or is it boilerplate? You can label it yourself in an afternoon, a regular expression will do embarrassingly well on it, and it is currently done by an LLM. That is the whole course in one example.

By the end you can

  • Decide between rules, a classical model and an LLM for a stated problem, and defend it with numbers
  • Produce and read the classification metrics that decision rests on, including calibration
  • Set a confidence threshold for human review from data rather than by feel, and state what it costs
  • Build a model selection matrix, and a build-versus-buy case a finance-literate reader would accept

0 of 24 required items complete

0m of 32h

M1 · When not to use an LLM

0/5 · 7h 15m

M2 · Metrics that make the decision honest

0/6 · 6h 30m

M3 · Human-in-the-loop

0/5 · 7h

M4 · Selection, build and buy

0/4 · 5h 30m

  • reading · 1h · tier 0 self-marked

    AI Engineering

  • optional

    reading · 45m · tier 0 self-marked

    Models overview

  • assignment · 2h · tier 2 panel-assessed

  • assignment · 1h 30m · tier 2 panel-assessed

  • retention · 15m · tier 1 machine-verified

M5 · Course project

0/4 · 6h 30m

  • project · 2h 45m · tier 3 artifact

  • project · 1h 30m · tier 2 panel-assessed

  • project · 1h 15m · tier 3 artifact

  • defense · 1h · tier 4 defended