Back to Projects

// Personal project

HackFusion

I built a hackathon team-matching platform to solve a problem I saw directly: participants struggling to find the right teammates through scattered Discord servers and WhatsApp groups.

Tech Stack

ReactTailwind CSSSentence EmbeddingsRetrieval-Augmented GenerationAI-Graded Live Interviews

How it works

The core challenge was matching people on what they actually know, not just keywords on a profile.

  • Built a matching engine using sentence embeddings to compare the actual meaning of a user's skills against a team's needs, instead of simple keyword matching.
  • Built an AI-led live skill-verification interview, where a candidate answers AI-generated spoken questions on webcam that get graded automatically to award a verified skill badge.
  • Built a document-chat feature powered by a retrieval-based system, so users can ask questions directly against uploaded documents like a hackathon's rulebook.

Why I built it

Hackathon participants were finding teammates through scattered Discord servers and WhatsApp groups with no real way to verify anyone's claimed skills, which made team formation slow and unreliable.

How I approached it

I used sentence embeddings instead of keyword matching so the engine compares actual meaning, added a live AI-graded interview to verify claimed skills instead of trusting a profile, and built a RAG-based document chat for hackathon-specific questions.

HackFusion screenshot 1
HackFusion screenshot 2
HackFusion screenshot 3
HackFusion screenshot 4
HackFusion screenshot 5
HackFusion screenshot 6
HackFusion screenshot 7