Data Scientist (GenAI) at Gestamp
A multi-agent system on Semantic Kernel, with Text-to-SQL under row-level access control and a golden-dataset evaluation, served as a streaming FastAPI service on AKS over hybrid RAG via Azure AI Search.
Two years at Gestamp, March 2024 to April 2026, as a data scientist working on GenAI. The central piece was a multi-agent system built on Semantic Kernel, including Text-to-SQL with RBAC impersonation — queries run with the caller’s permissions — and a golden-dataset evaluation to keep answers honest.
It was served as a streaming FastAPI service on Azure Kubernetes Service, with hybrid retrieval over Azure AI Search.
outcomes
- Built a multi-agent system on Semantic Kernel.
- Implemented Text-to-SQL with RBAC impersonation and golden-dataset evaluation.
- Served the system as a streaming FastAPI service on Azure Kubernetes Service (AKS).
- Used hybrid RAG over Azure AI Search.