World model research & training
On-siteAi, Borno State, Federal Republic of Nigeria
Job Summary
Design, train, and iterate on world models capturing real-world physical dynamics and robot-environment interactions using scalable pipelines built from real and synthetic datasets. Research novel architectures including diffusion-based and autoregressive models to improve fidelity, generalisability, and inference efficiency while developing methods for high-fidelity 3D scene generation via NeRF and Gaussian splatting. Curate training datasets combining robot fleet data, simulation outputs, and internet-scale visual assets, then define rigorous evaluation frameworks to measure sim-to-real transfer and close the gap between simulation and deployment. Mentor junior researchers, shape the long-term research agenda for world models and 3D generation, and publish findings at top-tier venues like CoRL and NeurIPS.
Required Qualifications
- Experience designing, training, and iterating on world models that capture real-world physical dynamics, robot-environment interactions, and long-horizon scene evolution
- Experience developing scalable training pipelines for world models using large-scale real and synthetic datasets
- Experience researching and implementing novel architectures including diffusion-based, autoregressive, and latent space models
- Experience investigating and improving model conditioning strategies for generalisation across diverse environments, robot embodiments, and task types
- Experience driving experiments in predictive world modelling including multi-step rollouts, uncertainty estimation, and sim-to-real transfer
- Experience researching and developing methods for generating high-fidelity 3D scenes including geometry, texture, lighting, and object placement
- Experience exploring and implementing neural scene representations such as NeRF, Gaussian splatting, and implicit surface models
- Experience building pipelines that automatically generate diverse, physically plausible simulation environments from minimal inputs
- Experience developing techniques for procedural and generative scene variation to support large-scale data augmentation and robot training
- Experience ensuring generated 3D scenes meet physical accuracy requirements for robot policy training and evaluation
- Experience designing and curating training datasets combining real-world robot data, synthetic simulation data, and internet-scale visual data
- Experience developing data collection, annotation, and preprocessing pipelines that scale with fleet size and simulation throughput
- Experience working with engineering teams to productionise training workflows on large GPU clusters
- Experience defining and implementing data quality standards and filtering strategies to maximise signal and minimise distribution mismatch
- Experience defining rigorous evaluation frameworks and benchmarks for world model fidelity, scene generation quality, and downstream robot task performance
- Experience measuring and closing the sim-to-real gap to ensure policies and behaviours trained in simulation transfer reliably to real robot deployments
- Experience publishing and presenting research findings at top-tier venues such as CoRL, ICRA, NeurIPS, or CVPR
- Experience mentoring junior researchers and engineers
- Experience contributing to shaping long-term research agendas for world models and 3D generation
- Experience staying at the forefront of developments in generative AI, 3D vision, and embodied intelligence
- Ability to work closely with simulation engineering teams to ensure research outputs are productionised effectively
- Ability to collaborate with robotics engineers and application teams to identify high-value research problems grounded in real deployment challenges
- Ability to work in a collaborative and inclusive work environment
- Ability to take responsibility for own work and the products created
- Ability to make educated choices that benefit the organization
- Local to position location
- Not requiring visa sponsorship
Hiring someone like this?
Get your role in front of qualified candidates on Sorce.