Until it isn't. I spend my time on the systems underneath — provisioning, deployment, and figuring out why a build failed, before it becomes someone else's problem.
“If a library is adopted by 200 teams, it's usually because it saved them a real afternoon.”
A Java build optimizer that saves $23.8M a year in datacenter provisioning costs. A CDK library I built alone, in Java and TypeScript, now used by 200+ service teams. Health tracking for Local Zone rollouts across 500+ services, with build time down 40%.
And an AI root-cause system, built on Claude Opus: 200,000+ tasks, 10,000+ diagnoses, across 5+ regions, with resolution latency down 30%.
MicroStrategy — reliability for an enterprise AI BI chatbot used by 14,000+ companies, as a junior engineer on a 10-person team. A Python test suite that took QA story points from 60 to 3.
UVA Biocomplexity — LLM simulations on HPC clusters for MRSA transmission, and clustering pipelines on hospital antibiogram data.
AgentVal is my open-source project for benchmarking multi-step AI agents — FastAPI, CLI, and a Next.js 15 studio. Measure agents like software, not vibes.
University of Virginia, B.S. Computer Science, 2020–2024. GPA 3.7, Dean's List. AWS Certified Cloud Practitioner.
Python · TypeScript · Java · C++ · SQL · Bash · CDK · Lambda · EC2 · S3 · Docker · SLURM · PostgreSQL · DynamoDB · React · Next.js · LLM evaluation