Robot Learning Engineer

Snapshot The mission of 120ft Factory is to start the new western industrial revolution. We are building the essential infrastructure to close the "Physical AI Deployment Gap." Our core thesis is that making agentic AI work at scale requires pairing cutting-edge, stochastic generative models with strict, deterministic execution. Description As the Systems & Simulation Engineer, you are the architect of our digital twin platform and the stochastic, agentic layer that runs on top of it. You own the layer where three things meet: the deterministic supervisor that orchestrates the factory, the learned policies that drive manipulation and motion, and the live physical state of real robots in production. Your job is to design how all of that fits together, and to build the ML training infrastructure that fills the twin with competent policies. Working hand-in-hand with our Systems Engineering team (who define the safety-state execution contracts), you will design the architecture that connects complex, real-world assembly processes to LLM-driven task planning, humanoid locomotion, and semantic reasoning. Your work also minimizes the sim-to-real gap, ensuring our AI explores logic and learns within realistic contact dynamics before those policies are admitted to physical hardware. Responsibilities

• Architect and maintain closed-loop digital twins (using NVIDIA Isaac Lab/Sim or MuJoCo) that act as both generative orchestrators and physical training environments for VLA policies.
• Map physical and logical flows across the system, co-optimizing humanoid locomotion, skills for dexterous manipulation, and part routing to resolve deadlocks in a dense micro-factory space.
• Interface directly with the deterministic supervisor layer, ensuring that the stochastic skills and task graphs generated by your models cleanly map to the strict safety-state envelopes enforced on the shop floor.
• Pioneer cutting-edge AI and coding paradigms—integrating emerging World Models and LLM-driven agents for semantic scene understanding, zero-shot task planning, and emergent logic exploration.
• Write high-performance Python/C++ infrastructure to support imitation learning, reinforcement learning, and automated synthetic data generation pipelines.
• Tackle the 'sim-to-real' gap directly, modeling accurate contact dynamics and physical interactions to ensure policies survive in the reality of atoms.

Requirements

• Ph.D. or Master's degree in Robotics, AI, Computer Science, Engineering, or a related field.
• Programming skills in Python and C++, with hands-on experience in ROS2 and modern physics simulators (NVIDIA Isaac Sim/Lab, MuJoCo).
• Proven experience applying machine learning (Reinforcement Learning, Imitation Learning, Embodied AI, or World Models) to multi-robot systems, locomotion, or complex manipulation tasks.
• A demonstrated history of building complex digital twins, simulations, or automated orchestration architectures from the ground up.

Preferred Qualifications

• You view the simulation not just as a visualizer, but as a rigorous contact-dynamics training ground and a generative logic engine.
• Experience integrating vision-language or foundation models into robotic control loops (e.g., using LLMs for proactive human-robot collaboration or task reasoning).
• Familiarity with coordinating both mobile bases (locomotion) and upper-body arms (manipulation) within shared spatial constraints.
• You have practical 'battle scars' from sim-to-real transfer and dealing with the realities of deployed systems (latency, degraded channels, teleoperation fallback).
• You are a pragmatic builder obsessed with state-of-the-art AI, prioritizing rapid iteration over writing "perfect" code.

Compensation We offer a top-tier Stockholm salary combined with a significant, early-stage equity grant as part of the founding team. We believe in rewarding our core builders as true owners of the business.