Math-CS + Cognitive Science (ML/NC) · UC San Diego
Click to select a section
I'm a Math-CS and Cognitive Science double major at UC San Diego, building full-stack tools, ML pipelines, and multi-agent systems. Outside of class and code, I play racket sports, perform with LionDanceMe, and shoot photo/video for fun!
B.S. Mathematics-Computer Science
B.S. Cognitive Science Spec. Machine Learning and Neural Computation
Activities: Association of Computing Machinery, Society of Undergraduate Mathematics Students, Data Science Society, Triton Engineering Student Society, Badminton Club
A.S. Mathematics, Highest Honors
Activities: Student Tutor, Growth Sector Summer Bridge
Essenergy Inc
IDX Exchange
Visual Learning Lab @ UCSD
Data Science Society (DS3) @ UCSD, La Jolla, CA
The Econ Lab @ UCSD
UC San Diego Department of Mathematics
Growth Sector, San Francisco, CA
City College of San Francisco, San Francisco, CA
NASA Goddard Space Flight Center (Remote)
ACM @ UCSD
ACM @ UCSD
LionDanceMe, San Francisco, CA
Bert Hsu Academy, San Francisco, CA
Built at UC Berkeley AI Hackathon 3.0
Civic tech tool that uses Claude's vision capabilities to flag dangerous street intersections and auto-generate civic outreach pushing for fixes.
Engineered scraping pipelines to collect and transform 2,500+ courses into a structured SQL database with AST-backed representations. Designed a TypeScript AST-based prerequisite resolver parsing nested AND/OR logic and rendering dependency DAGs in D3.js. A CI/CD pipeline auto-scrapes and republishes new courses each quarter.
Built at DiamondHacks 2026
Multi-agent student trading platform where autonomous Fetch.ai agents greedily negotiate multi-item trades on behalf of students; integrated BrowserUse for real-time market pricing and confidence scoring. Personally built the FastAPI backend framework (REST + WebSocket controllers, Supabase PostgreSQL, AWS S3), the BrowserUse pricing features, and authentication.
Converts UCSD WebReg schedule PDFs into clean, importable
.ics calendar files, with a browser-based weekly schedule preview before
download. The FastAPI + PyMuPDF backend parses course data directly from the PDF, generates
recurring weekly lectures, skips UCSD holidays automatically, and adds midterms and finals
as one-off events. Vercel deploy still in progress.
Two-tower neural collaborative filtering model to predict songs through collaborative and content-based filtering, mapping user preferences and 900,000+ tracks into a shared latent space. Fine-tuned embedding and epoch size, achieving 5.2 precision@10, 0.84 ROC–AUC, and reduced validation loss by 30%.
Built at JacHacks SF 2026
Civic advocacy generator that turns a community's position on a local issue into a tactical organizing kit: plain-language issue briefs, procedural playbooks with source-backed facts, three recourse paths, targeting advice that flags mismatches between impacted and decision-making zones, precedent-based strategy, bilingual outreach packet.
Exploratory generative modeling project synthesizing Minecraft-style textures with diffusion models; currently investigating VQGAN/VQVAE approaches as a next step.
UCSD course browser combining WebReg++-style scheduling with ClassGraph's prerequisite visualization, backed by a custom scraper against UCSD's TSS OData API.
Vehicle routing optimizer for LionDanceMe operations, modeling multi-stop routing as a constrained Traveling Salesman Problem using road networks and travel-time matrices from OpenStreetMap data. Integrated OR-Tools to enforce time-window and overlap constraints across 40+ daily events, improving schedule feasibility and reducing total travel time.
Research Assistant, UCSD
Maintainer of vllpy, the lab's internal Python toolkit for multimodal and layer-specific embedding models. Added parallel and asynchronous batch processing capabilities.
Feel free to reach out below.
Coming Soon!