A closer look
Selected work.
Backend systems, the infrastructure around AI, and the evaluation that makes both better.
5 overviews
Execution reliability
Durable AI workflows
Designing and implementing the backend lifecycle around long-running AI work, with clear progress, recoverable state, and reliable access to completed results.
Model serving
Self-hosted inference
Building LLM serving and evaluating it against complete application workloads, with attention to tail latency, throughput, and GPU resource use.
Evaluation quality
Applied ML evaluation
Evaluating observational data with causal ML methods, careful feature validation, and explicit limits on what an estimate can establish.
Data correctness
Backend reliability & data correctness
Improving event-driven services through production debugging, database tuning, and recovery tooling that makes operational intent explicit.
Testing infrastructure
Agent-assisted testing
Designing a testing platform that combines repeatable regression checks with bounded browser agents and reviewable evidence.
Professional contributions are described at a general level. Company implementation details are omitted.