Every deployment teaches the next one. Your job is to make that loop turn faster and teach more each time.

Responsibilities

• Set the research agenda for Cortex: vision-language-action models, world models and planning, learning from interaction on real robots, and generalization across tasks and embodiments.
• Make the big bets: which directions get people and compute, and which ones we stop.
• Own the data strategy: the mix of deployment, teleoperation, open-source and synthetic data, and what the fleet collects next.
• Own the compute strategy, from large-scale training to real-time inference on the robot.
• Build the evaluation science: benchmarks that predict performance in a customer’s operation, not only on a leaderboard.
• Build and lead the research organisation in Zurich and Stuttgart: structure, leads, rhythm.
• Set the hiring bar and make Sereact the place the best people in robot learning want to join.
• Take research into production with engineering, so results reach the fleet in weeks and hold up in 24/7 operation.
• Decide with the founders what we publish, open-source or patent, and build our standing at CoRL, RSS, ICRA, NeurIPS and ICML.
• Be Sereact’s scientific voice with customers, partners, universities and investors, and build academic partnerships.
• Partner with the CTO, product and operations, so research, engineering and deployment pull in one direction.
  

Not in scope, by design

• Software platform, hardware and field deployment engineering. They stay with the CTO and operations: the CTO owns how we build and ship, you own what we learn next.
• A research lab beside the business. Research at Sereact ends on a robot, ideally at a customer.
• Product roadmap and pricing.
  

What makes this role hard

• Research and production compete for the same robots, people and compute. You make that trade-off every week, in the open.
• Cortex is not a blank page. It already runs in customer operations, and our founders are deeply technical. You raise the ceiling of a running system.
• A result counts when it holds up in a customer’s operation, not when the paper is accepted.
  

Your first 12 months

• First 90 days: know the fleet, the data and the team; agree the research roadmap with the founders; one evaluation standard everyone uses.
• By month 6: a research organisation with clear ownership and a release rhythm from experiment to fleet.
• By month 12: a new generation of Cortex in customer operations with measurable gains, key senior hires made, and work that puts Sereact on the map of robot learning.
  

What you bring

• You have led research in robot learning or embodied AI (VLA models, imitation or reinforcement learning, world models) and built a team around it.
• Your work runs on real robots, ideally outside the lab and for long stretches. You know what breaks first.
• Scientific standing: a strong record at the top venues and a network that great researchers want to join.
• You have made real calls on research bets, data and compute, and stopped what did not work.
• You explain science to engineers, customers and investors alike, and you enjoy it.
• PhD in machine learning, robotics or a related field, or an equivalent track record.

Nice to have: large-scale pretraining of foundation models · fleet learning or data engines · industrial manipulation · experience in both academia and industry.

Three questions we will ask you

1. Tell us about a model you took from a paper to a robot that ran for weeks without you. What broke first?
2. If you had our deployment data tomorrow, which bet would you make first, and what would you stop?
3. How do you decide what to publish and what to keep?
   

What we offer

• Robots in live production that learn every day, and the data that comes with them.
• A seat in the leadership team of a company scaling from Europe into the US, with meaningful equity.
• Large-scale compute and the freedom to build the research organisation you would want to work in.
• Teams in Stuttgart, Zurich and Boston; relocation support.
  

How we hire

Conversations with both founders, a research deep dive with the team, and a working session on a real problem from our deployments. We move fast.