Josh Garbutt

World Model Head Hunter  ·  This is Growth

Research talent for world models and embodied AI.

I work with research scientists and research engineers across multiple verticals within world models:

  • robotic manipulation
  • dexterous manipulation
  • locomotion
  • robotics
  • diffusion policies
  • vision-language-action models
  • world action models
  • RL / RLHF
  • video understanding
  • generative models
  • gaussian splatting
  • datasets
  • humanoids
  • embodied systems
  • simulations
  • agents
WeChat QR code for Josh Garbutt

scroll to look down ↓

Live jobs

Live jobs

A live, dated list of what I'm working on right now.

Research Scientist / Research Engineer

World models · Video generation

LondonZurichPalo Alto

Generative video steered in real time, built as a world model rather than a clip generator. Wants people who have trained large video diffusion or autoregressive models end to end and who care as much about controllability and latency as sample quality.

Research Scientist / Research Engineer

World models · Simulations

LondonZurichPalo Alto

Learned simulators that stay coherent over long rollouts and are used as training environments, not demos. Looking for researchers who have built or rigorously evaluated world models as simulators and who think about drift, physical plausibility and long-horizon consistency.

Research Scientist / Research Engineer

World models · Controllable / interactive video

Los Angeles

Real-time interactive video for game worlds, generated rather than authored. After engineers who can hold controllable generation at frame rate, coming from video models, inference optimisation or realtime graphics.

Research Scientist / Research Engineer

World models · Robotic manipulation

LondonZurichSan FranciscoPalo Alto

Manipulation policies running on real hardware, not just benchmarks. Wants researchers who have trained visuomotor policies and actually got them to transfer to physical robots, along with the data pipelines that make that repeatable.

Research Scientist / Research Engineer

World models · Diffusion policies

ZurichSan FranciscoPalo Alto

Diffusion-based action models for manipulation. Looking for people who have trained diffusion or flow-matching policies and can speak to inference cost, action chunking and how these behave once they leave simulation.

Research Scientist / Research Engineer

World models · Agents in gaming

San Francisco

Agents trained inside generated game environments. After researchers with real depth in RL or imitation learning who have trained agents at scale in simulated worlds and evaluated them beyond a single benchmark.

Infrastructure Engineer

Infrastructure · Large-scale GPU training

San FranciscoZurichLos AngelesUK

The training stack behind large video and world models: multi-thousand-GPU clusters, throughput, checkpointing and failure recovery. Wants engineers who have run distributed training at that scale in production and can say where the bottleneck actually was.

Research Scientist / Research Engineer

World models · RL / RLHF

LondonZurichSan FranciscoPalo Alto

Reward modelling and post-training across video, robotics and world models: RLHF, preference data, and RL on top of large pretrained models. Wants people who have run a full post-training loop in anger, not only PPO on a benchmark.

01 / 08

References

Candidates’ feedback

Contact

Worth a conversation?

Worth a chat if you're hiring in world models, or if you're a researcher thinking about your next move. Any of these reaches me, and I'd be glad to hear from you.