Hackathon Nest (AI Native Platform for Hackathons)

Why I Built This

When I moved to SF, I failed multiple hackathons. Zero feedback. Confidence shattered.

The worst part? I watched teams win by cramming every sponsor into Frankenstein apps while genuinely good projects got ignored. Someone adds Sentry.init() 30 minutes before deadline and wins the Sentry prize. Meanwhile, people who actually built something meaningful walk away with nothing.

The system is broken.

The Insight

I kept doing the same thing before every hackathon: stalking judges on LinkedIn, reading sponsor docs, trying to find what they actually want. It took hours.

Then it clicked—judges have public histories. Sponsors have clear needs. Themes have criteria. What if AI found where they all overlap?

Instead of guessing, you'd know exactly what to build.

What It Does

Research → Match → Generate → Build → Verify → Learn

The platform scrapes your hackathon's Luma page, researches judges' LinkedIn posts, analyzes sponsor products, and finds the intersection. Then it generates project ideas ranked by alignment score.

For team matching: fill a quick profile, get matched by complementary skills and working style. No awkward "anyone need a teammate?" for us introverts.

The real game-changer is the judging interface. Judges see the truth:

"CodeRabbit: DEEP (95%) — 12 files, 47 API calls" "Sentry: SHALLOW (23%) — added 37min before deadline, init() only"

Data doesn't lie.

And everyone gets feedback. Not just winners. 98% of participants normally walk away learning nothing. That's insane.

The Hard Parts

Building the overlap algorithm—balancing judge preferences, sponsor needs, and theme constraints—was genuinely difficult. Three-dimensional optimization with fuzzy inputs.

Also learned that "deep integration" means different things to different people. So I built verification that shows exactly how sponsors were used. No more subjectivity.