HackATeam

• 

  The results page

• 

• 

🌟 Project Title: SCEHacks Matchmaker 🔥 Inspiration

Hackathons are about collaboration, but forming a balanced team quickly is often chaotic. People cluster with friends or those with similar skills, which leads to overlap (like 3 back-end devs and no designer). We wanted to fix that by building an AI-powered teammate matcher that promotes complementarity over similarity.

💡 What it does

Smart Profile System: Participants enter their roles, skills, interests, and availability.

AI Scoring Engine: Powered by Google’s Gemini, it evaluates not just overlap but how complementary two people are (e.g., pairing a front-end dev with a back-end dev).

Instant Team Recommendations: Our algorithm returns ranked matches with clear explanations (“Strong complement: FE + BE, overlapping interests in AI”).

Diverse Data Seeding: The platform comes preloaded with 50+ diverse profiles across hackathon variants, ensuring broad test coverage.

Frontend + Backend Integration: A TypeScript + Vite frontend (deployed on Vercel) communicates with a FastAPI backend (deployed on Render) through a secure API.

⚙️ How we built it

Backend:

FastAPI + Uvicorn for API endpoints.

MongoDB for participant storage.

Gemini for AI-powered scoring, with strict prompts enforcing complementarity.

Guardrails + scoring rules ensure balance (overlap = capped, complement = boosted).

Frontend:

React (Vite + TypeScript) for fast development.

Tailwind CSS + ShadCN UI for a modern, clean design.

Recharts for simple data visualization of matches.

Vercel deployment with environment variables connecting to the backend.

DevOps:

Render for backend hosting with .env management.

Vercel for frontend with VITE_API_URL linked to Render.

GitHub for version control.

🚀 Challenges we ran into

TypeScript Type Mismatches: Our Match types drifted between API and frontend storage — strict TypeScript rules on Vercel caused multiple build failures.

Submodule Accident: Accidentally turned Backend/ into a Git submodule and had to reset Git history to push properly.

CORS & Deployment: Making Vercel frontend talk to Render backend required fine-tuned CORS settings and environment variables.

Prompt Engineering: Ensuring Gemini always returns strict JSON and prioritizes complementarity instead of similarity.

🏆 Accomplishments we’re proud of

Built a full-stack AI app in one weekend that runs live on the web.

Created a system that fixes hackathon team-forming pains we’ve personally experienced.

Learned to debug tough Git, TypeScript, and deployment issues under time pressure.

A sleek UI + working AI backend integrated successfully.

📚 What we learned

The importance of prompt design + guardrails when relying on LLMs.

Real-world TypeScript challenges with shared interfaces.

How to set up Render + Vercel + MongoDB for production-ready deployment.

That great teams aren’t built on similarity — they thrive on diverse skill sets.

🔮 What’s next

Add team size recommendations (e.g., “You + 3 others make a balanced team”).

Support multi-criteria matching (technical + non-technical traits).

Build an event organizer dashboard for monitoring team formation.

Integrate real-time chat or Discord/Slack bots for instant team building.

Try it out GitHub:https://github.com/amcrjlnv/SCEHacksFriSep5

Backend: https://scehacks-2.onrender.com/

Frontend: https://scehacks.vercel.app

Built With

• fastapi
• gemini
• html/css
• javascript
• python