Senior AI-Engineer / Agent-Architekt (m/w/d)

We are looking for an engineer who builds our AI-Agent platform for enterprise customers in regulated industries.

No tutorial LLM, no vibe engineering, no demo RAGs that fall apart at 10k embeddings. We build explainable AI with audit lineage, decision provenance, and pgvector at scale, because our customers want to not only explain their KI outputs but must. Investor-backed, Munich region.

Aufgaben

What you'll be building

• Production-grade AI-Agent platform with PydanticAI, LiteLLM, Langfuse
• RAG pipelines with pgvector at scale (Multi-Tenant Isolation, Index Rebuild Strategy, Hybrid Search)
• Multi-Tenant Token-Quota Service plus Eval Frameworks per Tenant
• MCP (Model Context Protocol) Tool Integration as First-Class Citizen
• Explainable Outputs, seamless Auditability from Prompt to Output for regulated industries
• Architecture decisions that hold in the banking-, insurance-, and pharmaceutical context

How we work

• Eval-driven, not vibe-driven. Every Agent Change goes through an Eval Suite.
• ADRs for every relevant Architecture Decision, not buried in a Slack Thread
• LLM Calls are measured, not guessed - Langfuse from day one
• Re-Ranking, pgvector-Sizing, and Retrieval Quality are empirically decided
• Direct line to the Founding Team, no layer of Project Managers in between

Qualifikation

What we're looking for

• 5+ years Python Production Engineering
• Earnest Production LLM Experience with Tokens, Latency, Cost, and Eval as daily reality
• RAG Engineering with pgvector-Sizing and Re-Ranking, not just