← Все заказы

Deep Learning Engineer

ChoicyWork

13 ч назад

Текст заказа

This role is shared by CHOICY – your go-to recruitment partner for top tech & product teams.
We’re helping one of our partner companies grow their team with someone awesome (maybe you?). You’ll go through us for the hiring process — we’ll keep things fast, clear, and friendly all the way through.
We know the team behind this product, and we’ll make sure you’re fully prepped for every step.
How to apply
Please complete and submit the application form
About the Company
The company builds commercially scalable, safe robots that work alongside people and extend what they can do. Its robotic platform was developed quickly and is already deployed in real industrial environments. The team is growing to take the product further.
About the Role
As a Research Engineer on the World Models team, the engineer will build action-conditioned generative models that predict how the world around the company’s robots evolves, from past observations and actions. The models predict future video, proprioception, contacts, and outcomes.
World models serve four purposes in the company’s stack:
a pretrained, physics-aware prior for VLA policies; an engine for rare data collection, cross-platform transfer, and sim-to-real transfer; a testbed for evaluating and testing policies before they run on hardware; a future-prediction rollout engine that shows what policies intend to do, for safety and planning. This is a hands-on individual contributor role. The engineer will design architectures, run large training jobs, and validate models against real fleet data from industrial deployments.
Responsibilities
Design and train multimodal world models (video, state, action, and language) using diffusion-based and transformer architectures. Build action-conditioned video prediction and dynamics models. These models need to stay physically consistent over long horizons, including in contact-rich manipulation, and serve as pretrained priors for VLA policies. Develop learned-simulator evaluation. This means scoring candidate policies offline, predicting real-world success rates before deployment, and rolling out policy futures to expose intended behaviour for safety review and planning. Generate synthetic rollouts and counterfactual experience to augment policy training, and measure their effect on downstream task performance. This covers rare events, cross-platform transfer, and sim-to-real transfer. Establish fidelity metrics and calibration protocols that show where the world model can be trusted and where it diverges from reality. Build data pipelines that turn fleet telemetry, teleoperation logs, and internet-scale video into training corpora for world models. Run scaling and ablation studies on architecture, data mixture, and context length, and communicate the findings concisely. Collaborate with the pretraining, RL, and manipulation teams to integrate world models into policy training and evaluation loops.
Requirements
Track record of training large generative models (video, world, or multimodal), with shipped models or published artifacts to show for it. Deep hands-on experience with modern generative architectures: diffusion models, autoregressive transformers, latent-variable models, or video prediction. Experience with large-scale distributed training: streaming datasets, checkpointing and state management, debugging numerics and training instabilities. Strong Python and PyTorch/JAX skills. Ability to profile kernels, optimize data loaders, and write maintainable research code. Empirical rigor: designing careful evaluations, running honest baselines, and documenting experiments clearly. Interest in grounding generative models in physical reality rather than pixels alone.
Nice to have
Experience with world models for robotics or autonomous driving (e.g. action-conditioned video models, learned simulators, model-based RL). Familiarity with robotics simulators (Isaac Sim, MuJoCo) and sim-to-real considerations. Experience using world models for policy evaluation or synthetic data generation at scale. Publications at top-tier deep learning conferences (NeurIPS, ICML, ICLR, CoRL, CVPR) or equivalent open-source contributions. Experience optimizing generative models for fast inference.
What the Company Offers
Competitive equity: stock options with meaningful upside as the company scales. 30+ paid days off: 23 days of annual leave, all UK bank holidays, and additional company closure days (including a Christmas–New Year shutdown). Private healthcare, including virtual and in-person care. Pension scheme with 8% total contribution (5% employee, 3% employer) on full earnings. Free daily breakfast, catered lunch and snacks in the office. Daily collaboration with strong engineers, researchers and product experts building the next generation of AI and robotics. Direct access to the founding leadership, real input on product direction, and the ability to drive key initiatives from day one.

Похожие заказы