01
Lead with the problem and the constraint
A tech-stack list is inventory, not evidence. Open with what needed solving and what limited the solution — latency budget, data volume, deadline, team size. Constraints are what make a decision interesting.
02
Show one decision in full
Pick a single trade-off, present the options you weighed, and state why you chose one. A single well-argued decision demonstrates more capability than six shallow project summaries.
03
Report outcomes with numbers
Before and after, with units. p95 latency, build time, error rate, cost per month. Numbers turn a claim into a result, and they invite a good follow-up question in an interview.
04
Say what you would change
A short 'what I would do differently' section reads as maturity, not weakness. It is the closest a written artefact gets to demonstrating how you handle feedback.