At Nordstrom, I architected a Python service that lets merchandising analysts ask questions about catalog and sales data using approved LLM tools. I designed agent execution with tool calling, structured-output validation, request limits, and trace records.
I built a RAG path over product and policy documents that returns source references, and added an evaluation harness covering 180 common questions before model or prompt releases. I also developed a Go Model Gateway and packaged the AI services for Kubernetes deployments.
At Wise, I designed Python services for payment operations and built a Go service that coordinated internal payment checks, processing roughly 8,000 requests per minute during peak traffic. I added Kafka event consumers, Redis-backed caching, and integration tests for payment workflows.
Earlier, I built Java services for loan applications at Credistar and parcel operations at InPost. Across those roles, I worked on Kafka event processing, SQL performance, REST APIs, and production support.
