AI Engineer (m/f/d)

We’re looking for an engineer who builds AI systems that create real impact in production (not just demos) and has worked with retrieval or agent-based approaches on enterprise data. At Natuvion, we move business-critical data and processes between platforms. This data enables powerful AI — but also comes with complexity, sensitivity, and high stakes. We use AI to automate internal processes and to build intelligent features into our products, and you’ll work across both areas. We operate with strictly protected data that is never exposed to external LLMs, supported by a strong authorization model. Self-hosted models are an important part of our strategy. Our core stack is built in Go with gRPC-based services, including a custom AI router. While AI work may start in Python, production services are built in Go — and you’ll own them end-to-end. We deliberately stay flexible with our AI stack. The field evolves quickly, and we value engineers who stay current, think critically, and bring ideas to the table. This role focuses on building robust, production-grade systems designed to be operated, evaluated, and improved over time. We’re looking for someone who prioritizes maintainability, thoughtful evaluation, and chooses the right tool for the job. If that excites you, we’d love to hear from you. Your Responsibilities: Design, build, and operate RAG solutions across the full spectrum — from semantic search through to retrieval over highly structured data and knowledge graphs Build agentic systems that perform multi-step tasks reliably and safely against enterprise data Drive automation of internal processes with AI, and ship AI features into customer-facing products Own retrieval quality end to end: chunking and embedding strategies, hybrid and re-ranking approaches, and the evaluation that proves it works Evaluate and recommend the tooling for our AI stack — vector stores, orchestration, retrieval frameworks, model choices — and drive those decisions together with our architect. We expect informed proposals, not a wait for instructions Author production services in Go (gRPC) and integrate them with our AI router and existing platform; prototype and iterate in Python Make grounded technical decisions — retrieval vs. fine-tuning vs. prompting vs. a classical approach — and defend the trade-offs in architectural reviews