About this role
About the RoleThis is a founding ML engineering role at an early-stage AI startup, sitting at the intersection of post-training, agent systems, and research infrastructure. You will own the ML function from day one, working closely with the founding team to define technical direction, run experiments, and build the systems that power the next generation of AI agents.
What You'll DoStructure, filter, and score experimental trajectories to build high-quality training data pipelines.
Design and implement evaluation frameworks and benchmarks that measure model reasoning, planning, and improvement.
Build reliable agent environments, tool interfaces, observability systems, and replay infrastructure.
Establish robust validation and provenance tracking for trajectory and data quality.
Set ML roadmap priorities across systems, experiments, and hiring decisions.
Lead and grow the ML team as the company scales.
What We're Looking For3 or more years of machine learning engineering experience delivering production ML systems.
Hands-on experience with post-training data pipelines, including structuring, filtering, and scoring training data.
Direct experience designing and implementing evals or benchmarks for ML models.
Strong Python skills and comfort with systems-level programming for ML infrastructure.
Deep understanding of trajectory data, reward modeling, and agent decision-making systems.
Experience with reinforcement learning algorithm implementation and tuning.
Experience building replay systems, debugging tools, and observability systems for agent trajectories.
High agency, comfort with ambiguity, and the ability to bridge research and production seamlessly.
Background at a frontier AI lab or on a post-training or evals team is a strong plus.
Compensation & BenefitsSalary range: $100,000 to $200,000 USD annually. Visa sponsorship is not available.
LocationPrimary location is Munich, Germany, with additional offices in Zurich and San Francisco. This is an on-site role; remote arrangements may be discussed on a case-by-case basis.