Building systems
that run, not just demo.

I'd rather ship one project that survives a
code review than five that don't.

Two years into writing code, and most of that time went into learning the difference between a working demo and a working system. Most student portfolios optimize for looking impressive. Mine optimizes for surviving scrutiny a recruiter who actually clones the repo, reads the README, and checks whether the numbers add up. That means synthetic data with documented provenance instead of unverifiable claims, incremental commit history instead of a single dropped-in blob, and architecture that matches what I call it a "multi-agent system" that actually passes data between agents, not four parallel one-shot calls wearing a label.