Senior AI Agent Engineer

Company Description

a confidential company - Building the first agentic AI network for finance operations

Job Description

You will own the architecture of a multi-agent system designed to automate $90 trillion in manual business transactions. You will design how autonomous agents communicate, process financial documents, and navigate business logic. This is a foundational role where you define the playbook for agentic systems in a zero-trust, high-stakes financial environment.

Location

Munich, Germany

Why this role is remarkable

• You will join a founding team that has built 10 startups, raised over $400M, and created 2 unicorns, offering rare access to elite venture-building expertise.
• You will lead the transition to AI-first engineering, orchestrating parallel coding agents with a significant budget while focusing your time on high-level system design and architectural rigor.
• The role offers deep technical ownership of a compounding data moat where every company joining the network makes your autonomous agents smarter across the entire supply chain.

What You Will Do

• Design complex agent boundaries and orchestration patterns, maintaining the critical separation between LLM reasoning and deterministic rule enforcement.
• Build sophisticated document processing pipelines using LLM-based parsing and vision models to extract structured data from invoices and contracts.
• Develop a Neo4j knowledge graph using GraphRAG patterns to capture implicit business knowledge and the "why" behind financial decisions.

The ideal candidate

• Has 4-7 years of production experience with Python or TypeScript, with a proven track record of designing event-driven systems or data pipelines.
• Possesses hands-on experience with agent frameworks like LangGraph or LangChain and a strong architectural opinion on when to use LLMs versus rules engines.
• Demonstrates a security-first mindset regarding untrusted input and prompt injection, alongside a passion for the evolving landscape of AI-powered financial workflows.