Three internships, three real problems
Each case covers the actual problem, what I decided, and what came of it — not just a list of what I built.
FlyRank
AI-Assisted Workflow Drill
Built the same settings form twice — lazy prompt vs. precise spec — to prove that directing AI carefully changes the actual result, not just the process.
Read the case →DecodeLabsCode Review & Bug-Fix Tool
A Python tool that doesn't just flag bugs — it explains why they happened, helps refactor, and remembers past fixes for next time.
Read the case →ElevvoSentiment Analysis on Movie Reviews
Trained and compared three models on 50,000 IMDb reviews — then found the model's real failure mode wasn't what the textbook says it should be.
Read the case →