Ian Arredondo

I am a Computer Science student at Texas A&M University San Antonio finishing my degree in December 2026. I build data systems that solve real problems, two of them running in production right now. My work spans data pipeline engineering, analytics and BI, AI and RAG systems, and process automation. Based in San Antonio, open to opportunities across Texas.

SwiftLot is a live auction intelligence platform built on one of the largest salvage vehicle auction networks in the country. What started as a tool to help my dad navigate car auctions grew into a production system that has tracked over 27,000 vehicle sales nationwide since February 2026, now serving 1,000+ unique visitors a month. The platform aggregates real time bid data via server sent events, enriches Texas vehicles with odometer history from state inspection records, and archives final sale prices to a growing PostgreSQL database at auction close. The visualizations below are pulled straight from that live database, showing where the auctions happen, what's actually selling, and how price moves with volume week to week.

OnTimeSA answers a question a bus schedule can't: how reliably do San Antonio's buses actually run? It polls VIA Metropolitan Transit's live vehicle feeds every two minutes and compares them against the published timetable across 89 routes and 6,097 stops, building a record of what riders actually experience. The finding that surprised me most was that the worst performing routes keep schedule barely half the time, and the timetable hides it, because trips running early and trips running late cancel out in an average. It also detects bus bunching, where two buses scheduled fifteen minutes apart arrive within a minute of each other and leave the gap before them unserved. Riders can look up their exact boarding and exit stops to see how that specific segment has actually performed.

JagUnify started as a capstone project built around a real problem. New students, especially international students, had no reliable place to get answers about courses, admissions, or financial aid, and TAMUSA's own catalog site often left them at dead ends. Working with a team of five, I built an AI powered academic advisor that answers those questions with citations back to the official catalog, and refuses when it can't back one up. It's grounded in 2,028 indexed records across 30 plus departments, retrieved through vector search and cross encoder reranking, and evaluated at 100% accuracy across a 20 case test suite. I owned the backend, data pipeline, retrieval system, and the Power BI analytics below, which break the knowledge base down by department and level.