Production RAG · Measured & regression-gated · EU-sovereign
I build RAG and GenAI systems held to a production bar: retrieval measured against an expert-validated gold set, regression-gated, deterministic and auditable by design. When the data can't leave European jurisdiction, the stack stays sovereign end to end — self-hosted models, EU infrastructure, no US service in the path.
Work
End-to-end knowledge and retrieval layer for a production RAG assistant — a deterministic, rule-based ingestion and chunking pipeline (no ML, no LLM), structure-aware indexing for AWS Bedrock, two-stage retrieval with reranking, and a four-pillar evaluation harness.
A controllable, EU-sovereign environment for agentic coding: LiteLLM tier-routing across on-device and EU-managed providers (any OpenAI-compatible backend swappable in), an adversarial reasoning/execution gate, and a single gateway guardrail.
A spec-driven engine that generates SAP-HANA warehouse objects from YAML and parses existing ones back — bound by a shared intermediate representation and byte-stable roundtrip tests.
A deterministic framework that makes AI-agent software delivery reproducible, auditable, and approval-gated — a no-LLM kernel, single-shot agents, enforced invariants, and a tamper-evident audit log.
Let's talk
Whether you need a RAG system proven correct and kept that way as it changes, or a sovereignty constraint solved end to end — or you're a delivery partner who needs someone who has actually shipped both — happy to talk systems. No pitch needed.
Building AI that has to be provably correct — or that has to stay in-jurisdiction? Get in touch.
General information, not legal advice · see the eval