CourseAI Engineering: The Complete Course · Module 6: Transformers · part 48 of 105
Part 48 · Module 6: Transformers

Lab: Decide: prompt, retrieve or fine-tune

16 min read·13 Oct 2026
Module 6 · Lab · About 60 minutes · Needs: lessons 3 and 4 (the three rungs, the change test, the cost columns)

What you are building

WHAT THE LAB PRODUCES 10 questions 3 rungs scorecard change test decision with key strings prompt, retrieve, fine-tune score, tokens, ms two lines edited four lines, checked
The finished piece: ten policy questions run through three rungs (blue, your pipeline), scored and timed, then rerun after a policy change (amber, the uncertain part), ending in a decision record that a check enforces (green).

Module 6 ends with a written decision for Priya, Dev and Marta: should BrightCart's assistant answer policy questions by prompting a hosted model, by retrieving the policy line, or by fine-tuning a small model of its own? Lesson 4 made the argument with one question. A decision needs more. In this lab you run ten policy questions through all three rungs, score every answer, record tokens and latency, then change two policies and run again. The product is a scorecard with real numbers and a filled-in decision record. Nothing here is new machinery: the rungs, the fine-tune loop and the cost table are lesson 4's. What is new is doing it ten times, measuring, and writing it down.

The rest of this course is yours to keep

This course is bought on its own, once, and stays readable afterwards, including the parts added to it later.

Buy this course · ₹399See what is inside