At Morgan Stanley, I architect and lead production multi-agent LLM platforms that answer prime-brokerage risk, exposure, and margin questions for global risk teams.
I've led graph-based and supervisor-based agent architectures, with evaluation harnesses, full request tracing, guardrails, caching, and regulatory audit logging across multiple regions. I also delivered a centralized REST/gRPC platform processing roughly 8,200 records daily and netting-group management for thousands of accounts.
My foundation is in Java, Spring Boot, Kafka, and event-driven microservices, complemented by Python and agentic AI tooling. Previously at American Express, I co-designed a SARIMAX forecasting service that reduced forecast turnaround from hours to about 15 minutes.
