Software Engineer II
Zenoti · Aug 2024 – Present
- Architected and built end-to-end an agentic RAG platform (LangChain, Claude, PGVector, AKS) turning natural-language questions into governed Redshift SQL scaled to production beta across approximately 100 organizations, 1,000+ users, and 3,000+ queries/day.
- Designed a 5-agent pipeline (intent classification, schema lookup, clarification, SQL generation, output/visualization) via LangChain, backed by a custom Python SQL validation gate enforcing query safety pre-execution.
- Built a PGVector-based RAG layer embedding KPI definitions to retrieve the most relevant metrics per question, grounding LLM-generated SQL in correct business semantics.
- Implemented a self-correcting retry loop (up to 3 attempts) feeding SQL errors back to the LLM, plus guardrails for prompt-injection defense, SQL query obfuscation, and multi-tenant row-level access control.
- Architected and delivered 4 net-new independently deployable microservices in C#/.NET (ASP.NET Core) within a 200+ engineer platform, containerized with Docker and orchestrated on AKS reducing cross-team release conflicts by approximately 40%.
- Optimized SQL Server stored procedures via execution plan analysis and index tuning reducing average query execution time by approximately 35% for high-volume transactional reports.
