Behavioral & Portfolio

AI roles weight shipped projects heavily. This round is where the capstone projects pay off.

What interviewers probe on your projects

• Why these design choices? Model selection, context strategy, framework-or-not.
• What broke? Every real agent project has war stories — hallucinated tool calls, runaway loops, eval surprises. Have two ready with the fix.
• How did you measure it? "It seemed to work" is a red flag; a small eval suite is a green one.
• Cost & latency — did you think about tokens as money?

Story structure

Situation → constraint → decision → measured result → what you'd change now. Keep each project to 90 seconds unprompted; go deep when pulled.

Portfolio checklist

• 2–3 projects from Build These First, deployed or demo-able
• READMEs that show architecture diagrams and eval results
• One project with real users or real data, however small

Work in progress

This page is a scaffold — worked examples and depth are being added. Want to help? Grab a good first issue or read the contribution guide.