VLA Pre-training & World Models Engineer (Visa Sponsorship - London, UK)

Open to candidates worldwide. Visa sponsorship available. A well-funded humanoid robotics company with robots already running in industrial settings is hiring two senior deep learning engineers for its core vision-language-action model team. The team trains base models from scratch on multi-robot trajectory data and large-scale video, and the models must run in real time on the robot itself. The role is a senior individual contributor position with no management track, and results are shared publicly as engineering blog posts. Candidates can focus on either VLA pre-training or world models.

What you'll do

• Pre-train and fine-tune base vision-language-action policies on diverse multi-embodiment trajectory data, and post-train them with behaviour cloning and reinforcement learning.
• Work with the data collection team to define what good training data looks like, find failure modes, and keep data diverse and well covered.
• Build and maintain data pipelines for synthetic data, teleoperation logs and labelling, and automatically feed new failure cases back into retraining.
• For the world models track: design action-conditioned video and dynamics models using diffusion or transformer architectures, and use them for offline policy evaluation and synthetic rollouts.
• Scale distributed training with the MLOps and data platform teams and optimise models for real-time inference on edge hardware.
• Run scaling and ablation studies, and document experiments and trade-offs clearly.

What you need

• 3+ years of building deep learning systems in industry or research, with shipped models or published work.
• Hands-on experience with at least one of: LLMs, VLMs, or image/video generative models, including architecture, training and inference. Robotics experience is not required.
• Experience actually training models, including pre-training runs and distributed training infrastructure. API integration or light fine-tuning alone is not enough.
• Strong Python and PyTorch or JAX, with the ability to profile code and debug training instabilities.
• Strong coding fundamentals; the interview process includes a LeetCode-style coding round.
• Working English and the ability to work on-site in London.

How to apply

Get the free i'llTip app to see the employer's details and apply. i'llTip's AI can tailor your résumé to this job, and you'll get alerts for similar roles.

Prefer the website? Continue on illtip.comOr get the app: Android · iPhone

🔔 Get new jobs like this first: join Tech & AI Jobs on Telegram · or get personal job alerts