All projects

Quantum Quest

Quantum concepts are easier to understand when you can predict a circuit, run it, and compare the result with what you expected. Our four-person team built Quantum Quest during the 48-hour Qollab x IonQ Global Quantum Hackathon.

Tech highlights

• 21 runnable Qiskit lessons
• Visual circuit builder and noise comparison
• Break the Code error-correction challenge
• AI tutor with an offline fallback

GatherBite

Group food orders look simple until dietary needs, budget, serving size, and live menu data collide. GatherBite separates planning from approval and uses deterministic checks as the final gate.

Tech highlights

• Separate planner and verifier agents
• Deterministic constraint checks
• Fail-closed handling for missing evidence
• Cart verification before and after execution

Personal Finance Dashboard

Personal spending data is often scattered across accounts, making it hard to see where money actually goes. This dashboard pulls transactions into one place, groups them by category, and uses the OpenAI API to turn the numbers into plain-language insights.

Tech highlights

• Express/Node.js API
• OpenAI-powered insights
• Transaction categorization
• React dashboard UI

Free-tier backend — may take up to a minute to wake up if idle.

Real-Time Energy Dashboard

Energy usage data updates constantly, but it's hard to monitor or make sense of without a live view. This dashboard processes incoming readings in real time and visualizes consumption trends so unusual patterns are easy to spot.

Tech highlights

• FastAPI backend
• Real-time data pipeline
• Recharts visualizations
• Separate frontend/backend deploys

Free-tier backend — may take up to a minute to wake up if idle.

FastAPI Docs Q&A — Hybrid RAG

Technical docs mix exact terms with conceptual explanations, so search alone tends to miss one or the other. AI answers can also cite sources that don't really back up their claims. This system combines dense and keyword search with an independent fact-check on every citation, verified against a hand-written eval.

Tech highlights

• Hybrid dense + BM25 search
• Reciprocal Rank Fusion
• Citation fact-checking
• Evaluated on 18-question eval set
• Streamlit interface

Free-tier app — may show a "waking up" screen for about a minute if idle.

DevProof

GitHub profiles and resumes are easy to skim but hard to verify. DevProof pulls real evidence — commits, file structure, source code — and has an LLM score it against a fixed rubric instead of eyeballing it, for both individual repos and whole engineering teams.

Tech highlights

• GitHub OAuth sign-in
• LLM-graded reports on a fixed rubric
• Resume-vs-GitHub verification
• Org-wide contributor leaderboard
• FastAPI + Postgres backend

Sign in with GitHub to try it — free-tier backend may take a minute to wake up if idle.

Ultimate Tic-Tac-Toe — Daily Puzzle

Wanted to combine a genuinely deeper strategy game with a Wordle-style shared daily format, and needed a real AI opponent rather than random moves — so I wrote a minimax engine with alpha-beta pruning and a deterministic, backend-free daily seed.

Tech highlights

• Minimax AI with alpha-beta pruning
• Deterministic daily puzzle (Wordle-style)
• Vitest suite covering engine, AI, and puzzle generator
• Shareable emoji result summary
• No backend — static Vite build