Projects
Click any project to expand for details. Use the case‑study link for a dedicated page.
-
- Built full-stack web app (React, Node.js) with user auth, messaging, posting, and commenting.
- Implemented AI recommendation model that learns artist beat preferences.
- Ranked Top 10 in OutsideLLMs Hackathon (Official Hackathon of Outside Lands).
- Impact: Enables targeted beat submissions, bypassing industry gatekeeping.
-
- Developed voice-first AI agent using OpenAI Agent SDK, BrightData, NLX, and MongoDB.
- Logged meals, tracked nutrition, and generated custom workouts on demand.
- Placed 2nd in BrightData track & 3rd in NLX track (MCP AWS Enterprise Agents Challenge).
- Delivered working MVP within hackathon constraints, integrating multi-API data pipelines.
-
- Built full MCP Client + Server architecture for multi-step event planning.
- Automated vendor/venue communication, RSVP tracking, and calendar scheduling.
- Achieved Judges Award in MCP & Agents Hackathon (13 teams).
- Tech: Custom MCP protocols, React, Node.js, real-time dashboards.
-
- Scraped trending news, generated product ideas, designed assets, and deployed purchasable websites.
- Completed 0–100 build pipeline in under 2 hours using Buildship, Cloudflare Workers, and ElevenLabs voice tools.
- Won 1st Place at AI Hackathon (125 participants).
-
- Engineered $70 mobile sensor + Swift app replacing $10K+ commercial solutions.
- Led 25-person team, collecting 10K+ geotagged temperature points across Livermore.
- Adopted by 40+ students; informed city decisions on cooling infrastructure.
- Tech: Swift, Arduino, Data Science, ArcGIS.
-
- Built React Native app using Gemini API with 93% classification accuracy.
- Deployed in 2 schools (30+ users), cutting waste audit time by 2 hours/session.
- Won $1K sustainability scholarship; partnered with LVJUSD + StopWaste.
-
- Trained model on 8K biomedical samples to achieve 95% accuracy.
- Combined proteomic + genomic data using bioinformatics pipelines.
- Earned Director’s Award at regional science fair.
- Tech: Python, TensorFlow, bioinformatics tools.
-
- Programmed 3 CNNs on Edge Impulse for real-time detection/classification (85–95% accuracy).
- Deployed on Arduino hardware; completed pick–stack cycle in <90 seconds.
- Advanced to California Science Fair.