Projects
koding 4 kids
- Developed an interactive coding platform supporting 50+ concurrent users, powering the lessons page and the personalized AI tutor Blossom. Built and deployed API endpoints for AI-driven guidance, integrated an MCP server with the Claude model to generate dynamic learning paths, and implemented Redis caching with lazy loading and write-through to reduce AI response latency by 60%. Structured a SQL database to enable real-time progress tracking and efficient AI model inputs, improving lesson completion rates by 40%.
- Secured $10,000 in seed funding within three months and launched pilot programs in 2 schools and 1 library, expanding computer science education access by 30% in targeted Minnesota districts.
taylor.ai (hackharvard 2025)
- Built an AI-powered fashion crawler using Python (FastAPI) for backend orchestration and JavaScript for the interactive frontend, leveraging Google Gemini to convert natural-language prompts into optimized queries and SerpApi to fetch data from 30B+ fashion items in real time.
- Developed personalized outfit generation by mapping items onto user-uploaded photos with Gemini's Nano Banana model, delivering live pricing, product links, and achieving over 90% visual alignment accuracy.
fill.ai (hophacks 2025)
- Engineered multilingual, conversational form filler using Python and ElevenLabs TTS, performing vector-based image processing to generate forms in English.
- Conducted QA testing with 5+ international students during the hackathon, improving conversational accuracy and form completion speed by 35%, creating a seamless experience that makes English paperwork accessible without typing or translation.
