Software engineer
I build the systems that let AI agents run code, use tools, and work safely.
From model serving to the tools agents use. My work sits where software meets production constraints.
Amazon Bedrock AgentCore
I build Code Interpreter and Browser, and work on virtual-machine startup and restore. The focus is isolated execution that stays useful in production.
Read my experience on LinkedInAmazon Bedrock
I worked on Claude model launches, production hosting, and regional expansion. That work shaped how I think about reliability and capacity.
Read the project reflectionAmazon CloudWatch Logs
Earlier, I worked on the launch of cross-account observability. It is part of my background in distributed systems and cloud infrastructure.
View my backgroundNYU Shanghai
English + Español
I’m interested in what it takes to make capable agents dependable.
My background spans agent runtimes, sandboxing, model-serving infrastructure, and observability. Today, my focus is agent security.
I enjoy conversations about the practical details: what agents can access, how their code runs, and how we understand what happened.
Code, experiments, and a quantum-computing notebook collection.
Working on agent security, runtimes, or model-serving systems? I’d like to hear what you’re building.