Software Engineer II at Brain Station-23

Architected agentic commerce platforms that enable store owners to build and deploy custom AI agents through MCP tooling. Designed multilayer orchestration across 15 scenarios and built an MCP module with more than 32 native tools. Engineered feedback-driven LLMOps pipelines for evaluation dataset creation. Developed a multimodal AI shopping agent using .NET 9, C#, Python, Qdrant, OpenRouter, SQL Server, and Docker. Built a three-stage hybrid RAG pipeline using dense and sparse vector retrieval with reciprocal rank fusion, implemented live database rehydration for price and stock accuracy, and introduced incremental catalog re-embedding that reduced monthly inference costs by 10x. Led migration of 13+ nopCommerce modules from .NET 5 to .NET 9, removing 25,000 lines of dead code. Built an AI QA agent framework that reduced checkout testing time by approximately 75% and developed cache-busting middleware serving 30,000 requests per month.