At Glean, I ship LangChain-based LLM pipelines and stateful AI agents for enterprise research workflows, combining retrieval, structured outputs, tool execution, and internal API integrations.
I led a shared AI platform and model gateway across Anthropic Claude and OpenAI models on AWS Bedrock, Google Vertex AI, and Azure OpenAI. The platform supports routing, fallback, caching, retries, and observability and is used by six product teams.
I also fine-tune PyTorch transformer models for ranking, embeddings, and domain-specific representations, owning work from dataset construction through production monitoring. I established evaluation infrastructure with 500+ regression cases and automated release gates, while reducing AI/ML serving cost by 18% and p95 latency by 25% without lowering measured quality.
Before Glean, I shaped platform and infrastructure direction at HashiCorp, built distributed backend systems at Fastly, and developed event-ingestion and alerting APIs at PagerDuty. I bring more than a decade of experience in reliable APIs, control planes, production operations, and high-volume distributed systems.
