Blacksburg, VA
Mayank Goel
I am a computer science student at Virginia Tech, finishing in December 2026. I build software that runs in production and can be checked. Everything on this site is deployed or downloadable, each one has a demo recorded against the real system, and where there is a number I can tell you exactly how it was measured.
What I care about
I would rather be checkable than impressive. Every project here says plainly what is measured and what is deliberately unfinished, and I have deleted my own numbers when I could not reproduce them.
Most of what I have learned came from things breaking in ways tests did not catch. A cache that could not round-trip its own type. A model that stopped using its tools the moment I asked for structured output. An occupancy score that returned a confident number for a place nobody had reported on. Those are the stories I like telling.
Where I have worked
Software Engineering Intern
- Proved the IRS portal infeasible at 250 records per file against 45,000 annual 1099-Q records; team took my API path
- Found the missing Software Developer TCC that would have stopped the migration, cutting open questions from 7 to 4
- Specified the integration in C4 at three levels and found only one of four XML libraries did true XSD validation
- Built the JWKS the IRS API required from a CA-signed X.509 certificate, verified eight ways including modulus match
- Regression-tested 80+ Banner 9 Finance pages against a baseline, with PL/SQL on Oracle confirming data integrity
- Migrated 8 components from Bootstrap to Tailwind and caught Copilot mismapping the two spacing scales every patch
- Implemented WCAG 2.1 AA accessibility fixes across an Angular application for a state agency under ADA Title II
- Fixed Jest tests in a zoneless Angular setup, replacing fixture-stability waits with explicit async timer advancement
Connected and Automated Vehicles subteam
- Trained on CAV architecture and sensor fusion with VT's EcoCAR team, supporting an SAE Level 2 adaptive cruise feature
Undergraduate Research Assistant
- Built Google OAuth 2.0 and Canvas API authentication with role-differentiated login for 50+ students and faculty
- Led 5 researchers building a leaderboard service on Node.js and PostgreSQL, validated with 20+ pilot users
- Configured GitHub Actions workflows replacing a manual local setup, so contributors onboard without a maintainer
A few things I built
Marquee
WhatsApp-native AI product photography for Indian micro-sellers. Send a product photo on WhatsApp, get a brand-ready ad back in minutes