I'm building production LLM agent infrastructure at Blue Water Autonomy, including the core agent loop, tool execution registry, and memory and context management on GCP with PostgreSQL and Claude on Vertex AI.
I created eval systems that combine offline gates with online detector and judge evaluations written to production traces, improving regression detection and model reliability for real customer traffic. I also operationalized the platform with Datadog observability, tracing, structured logging, kill switches, and data-minimization features for compliance and auditability.
Before that, I led backend development for Valero Energy's industrial IoT systems, building Python services on Kubernetes and GCP that processed millions of telemetry data points. I designed REST APIs and event-driven microservices, strengthened tracing and safe rollout practices, and mentored junior engineers.
At ThredUP, I built backend services for high-volume e-commerce orders and inventory, using Python, PostgreSQL, Redis, and RabbitMQ. Across my work, I own systems from API specs and design docs through on-call response, incident learning, and reliable operation.
