At Quantagonia GmbH, I architect Aura’s production platform across GCP Cloud Run, Cloudflare, Google Vertex AI, and Vercel, including provider-agnostic Claude routing to meet EU data-residency, availability, latency, and cost requirements.
I built Aura’s shared agent runtime with Claude Agent SDK and FastAPI to coordinate context, persistent memory, structured outputs, and human review. I also established Langfuse tracing and end-to-end evaluations against customer-derived golden problems.
At Quantagonia, I led the architecture and delivery of DecisionAI, a multi-tenant AI platform on AWS. I created its agent runtime and evaluation platform, with quality checks and regression gates in CI.
Earlier, at Cobrainer GmbH, I built an internal RAG product for RFI/RFP responses and an analytics platform with a Bedrock-powered assistant. My earlier work spans site reliability, cloud security, and identity and access management.

