In my free time, I’m usually playing , hanging out with my friends, , or reading a book from my ever-growing reading list. Currently, I’m reading Urban Creatures (opens in new tab) by Sarah Gray.
Studying healthcare access gaps in Dr. Alexander Michels’ group using population and facility data. Implementing place-based spatial accessibility measures modeled on the PySAL access package, including floating catchment area and gravity models, building on PostGIS work from LineFinder.
Built a full-stack conversational data analysis app that answers natural-language questions across multiple CSVs with text and charts. Designed an LLM planning pipeline that validates model-generated plans before deterministic Pandas execution, with a regression test suite. Now contributing to a live agentic AI project.
Python
FastAPI
Next.js
Pandas
OpenAI API
Docker
July 2025July 2025July 2025to Aug 2026to Aug 2026to Aug 2026
OIT Support Desk Analyst
UT Dallas
Supported students, faculty, and staff across phone, email, chat, and remote channels, troubleshooting AWS remote access, Duo, Teams, PeopleSoft, Azure, and hardware and software issues.
An AI-powered SSH honeypot that won the Vultr track at HackTX 2026. A fake company server lured in real attackers, told bots, humans and AI agents apart, investigated what they left behind with live web search, and showed every intrusion as a creature on a 3D globe. It caught 4,900+ attacks from 20 countries over the weekend, including a live Kaiji botnet attempt. The site now replays the recorded attacks.
Mobile-first DART transit navigator with a custom RAPTOR-style routing engine, versioned GTFS schedule ingestion, walking-aware trip planning, and transfer-risk ranking, backed by 1.9M+ stop-time records in PostGIS.
A 3D galaxy of about 5,000 films where every star is a film, placed by how it feels instead of its genre. Plot, themes, and poster colors are embedded and laid out with UMAP, rendered with React Three Fiber on WebGPU, and vibe search runs fully in the browser with Transformers.js.
ML pipeline on 128K NBA shots predicting make probability, benchmarking five model families and explaining the model with SHAP. Deployed as an interactive dashboard with a shot predictor and an out-of-sample player shot-making leaderboard.