CS (B.S./M.S.) @ University of Maryland · SWE Intern @ Visa · ML research @ UMIACS
I build systems that ship with proof — benchmarks, fuzzers, and fault injection live next to the code they test.
- Building LLM-powered document-processing infrastructure at Visa (Java/Spring Boot)
- Co-first author, CVPR 2026 Workshop — multi-center COVID-19 classification from chest CT (paper)
| Project | What it is | Receipts |
|---|---|---|
| LazerBook | Limit-order-book matching engine (C++23) | 29.4M ops/sec at p99 < 250 ns; 385M libFuzzer executions under ASan/UBSan, zero crashes |
| lsm_kv | LSM-tree storage engine, from scratch (Rust) | ~648K writes/sec (64–111% of RocksDB on write-heavy YCSB); zero acknowledged writes lost under SIGKILL fault injection |
| TerpAdvisor | Schedule generator + course recommender for UMD (FastAPI/Next.js) | CSP solver with forward checking, interval-tree conflict detection, seven-factor recommendation engine over live PlanetTerp/umd.io data |
| Multi-Source COVID-19 Detection | 9-model heterogeneous ensemble for chest-CT classification (PyTorch) | 0.928 macro F1 across 4 hospital centers — CVPR 2026 Workshop |
Java · Python · Rust · C++ · Go · TypeScript — Spring Boot, FastAPI, PyTorch, PostgreSQL, Redis, Docker/Kubernetes, AWS