Computer Science Student at the University of South Florida
CS Student @ USF · Backend & Distributed Systems
I'm a CS student at USF finishing up in December 2026. Most of my work lives in the backend — distributed systems, APIs, the kind of stuff where correctness under load actually matters. I've shipped a payment integration processing 100K+ daily transactions, built a rate limiter that holds up at 7,500 req/sec, and deployed a full-stack tutoring platform that real people use. I like understanding why things break before they do.
Feb 2023 - Sep 2024 | Tashkent, Uzbekistan
Worked on the backend of MARTA, a point-of-sale system processing over 100K transactions daily. Integrated Visa and Mastercard SDKs in Java, tightened payment flows with JWT validation, and cut failed transactions by 15% across 1,000+ merchants. Also built an internal analytics dashboard that replaced a manual reporting process.
May 2023 - Present | Tampa, FL
Keep the lights on for 800+ rooms across campus — AV systems, hardware failures, the works. Got good at diagnosing problems fast and figuring out why they happened so they don't come back.
Built a full-stack tunneling tool that exposes localhost services over a public URL using persistent WebSocket connections — the same core idea as ngrok, built from scratch. The Go relay server handles concurrent client connections via goroutines and channels, and uses subdomain-based routing to map each client to its inbound requests. A custom request ID correlation protocol lets multiple in-flight requests share a single WebSocket without collisions. Measured ~7ms average round-trip latency locally. Also identified and documented a sequential processing bottleneck in the Python client — the kind of thing that matters when you're proxying more than one request at a time.
A token-bucket rate limiter built to actually work in a distributed environment — not just on one instance. Redis handles shared state across nodes, and Lua scripts make the quota checks atomic so there are no race conditions under concurrent load. Tested it with k6 to 7,500 req/sec; P95 latency stayed under 11ms. Prometheus and Grafana for observability.
A full-stack learning platform I built and shipped. Spring Boot backend, React frontend, JWT auth, deployed on the cloud. Integrated Claude's API for grammar feedback and lesson generation, added caching to keep API costs sane. Currently serving real users with 30+ lessons and 300+ exercises.
Trained a classification model on 1,000+ clinical records — age, cholesterol, ECG readings — to predict cardiovascular risk. Hit 95% accuracy and wrapped it in a Streamlit app so it's actually usable by someone who isn't running a Jupyter notebook.
Email: bsevaraqq@gmail.com
Phone: (656) 200-9172
Location: Tampa, FL