AI Engineer | Reinforcement Learning | Physical AI Company (Humanoids) | Munich

AI Futures; München; On-site; EUR 90,000 - 140,000; Full time

Job summary

Ai Futures is partnering with a fast-growing Physical AI company on a search for an AI Engineer to build reinforcement learning and inference platforms for humanoid robots. The role focuses on scaling RL systems across GPU clusters, and delivering low-latency model serving in cloud and on-prem environments.

The candidate will work with an experienced robotics team to take large models from training to real-world robots, with English as the working language.

Qualifications

• 3+ years building large-scale RL or post-training systems (distributed rollouts, actor-learner set-ups).
• Hands-on with policy optimization methods such as PPO or GRPO, and frameworks such as veRL, TRL, NeMo-RL or RLlib.
• Production inference with vLLM, SGLang or TensorRT-LLM.
• Distributed systems experience with Ray, Kubernetes or Slurm.
• Proficient in Python, PyTorch or JAX, CUDA.

Responsibilities

• Build scalable reinforcement learning systems to fine-tune large models across GPU clusters, simulation and physical robots.
• Own model serving in the cloud and on-prem, with a focus on latency and throughput.
• Develop tooling to run and evaluate AI agents that plan and act over long, multi-step tasks.

Skills

RL systems

Policy optimization

RL frameworks

Model inference

Distributed systems

Python

PyTorch

JAX

CUDA

Tools

Ray

Kubernetes

Slurm

TensorRT

veRL

TRL

NeMo-RL

RLlib

PyTorch

JAX

Job description

AI Engineer | Reinforcement Learning | Physical AI Company (Humanoids) | Munich

AI Futures are partnered with a fast-growing Physical AI company on a search for an AI Engineer to build its reinforcement learning and inference platform.

Our client is a well-funded European robotics business with an experienced founding team and serious long-term backing. They are building their core AI team now, and this hire will take large models from training through to real robots.

The Role

• Build scalable reinforcement learning systems to fine-tune large models across GPU clusters, simulation and physical robots
• Own model serving in the cloud and on-prem, with a focus on latency and throughput
• Develop tooling to run and evaluate AI agents that plan and act over long, multi-step tasks

The Candidate

• 3+ years building large-scale RL or post-training systems (distributed rollouts, actor-learner set-ups)
• Hands-on with policy optimisation methods such as PPO or GRPO, and frameworks such as veRL, TRL, NeMo-RL or RLlib
• Production inference with vLLM, SGLang or TensorRT-LLM
• Distributed systems experience with Ray, Kubernetes or Slurm
• Python, PyTorch or JAX, CUDA

Nice to have

• Robot learning in simulation and sim-to-real transfer
• On-device inference on NVIDIA Jetson-class hardware

English is the working language. Full-time and permanent, based in Munich, Germany.