Projects
I built LeetMind because I wanted coding practice that focused on the topics I was actually weak at. It chooses and generates problems from each learner's history, then checks every generated problem with authored tests, an independent solution, and 50 reproducible randomized inputs before publishing it. Python and C++ submissions run in isolated Docker containers without access to the network, hidden tests, or expected answers.
LiftLedger
I built LiftLedger to make workout and nutrition tracking feel less like data entry. Its React dashboard and Telegram assistant turn text, voice messages, and food photos into structured logs that users approve before saving. It also pulls in Hevy workout history to chart set counts and training trends across exercises and muscle groups.
Experience
Research Software Engineer
2026-08present
Texas A&M University
Built an interactive React and TypeScript platform for exploring U.S. beef supply-chain scenarios, replacing a spreadsheet prototype with a model of independent businesses, mortality, biological variation, inventory, market prices, and sensitivity analysis.
Machine Learning Research Engineer
2026-062026-08
Texas A&M University
Reduced a veterinary classifier's peak memory use by 74%, from 2.6 GB to 689 MB, while preserving 90.83% accuracy. Also built a Python and Django RAG pipeline with Voyage embeddings and Supabase pgvector to ground Gemini care plans in cited sources.
Full-Stack Developer
2026-01present
Built job-ranking and tailored application workflows for a platform with 200+ registered users, then made long-running application jobs recoverable with persisted snapshots, retries, deduplication, and coordinated cancellation.
Quantitative Developer
2026-022026-05
Maroon Fund
Added Jensen's alpha, OLS beta, and tracking error to strategy backtests using SPY data from a three-source fallback pipeline, then combined six separate charts into one interactive Plotly report.