Remote job
Research Scientist Intern
Job details
About this role
Role overview Investigate a focused research question connecting multimodal physical-world data with intelligent behavior, and relate the findings to practical robotics and autonomy workflows. With guidance from a research mentor, you’ll build reproducible experiments using real data; the project may focus on world modeling, hand motion, video understanding, or annotation systems.
Responsibilities - Define a research question with a mentor, reproduce relevant baselines, and set evaluation criteria before scaling experiments. - Explore world models that learn temporal structure and action-conditioned dynamics from multimodal trajectories. - Develop or assess hand motion tracking and reconstruction, including temporal consistency, occlusion, and object interaction. - Improve grounded video captioning, temporal segmentation, or annotation pipelines, including uncertainty checks and human review. - Prepare datasets and evaluation splits, run ablations, and analyse failure cases as well as aggregate results. - Contribute tested research code, experiment records, and technical reports; support publications or open-source releases when appropriate.
Requirements - Current MSc or PhD study in machine learning, robotics, computer vision, or a related field, or equivalent research experience. - Strong Python skills and practical experience training and evaluating models with a deep-learning framework such as PyTorch. - Depth in at least one area: generative or world models, 3D vision and tracking, vision-language models, or data-centric machine learning. - Careful experimental reasoning, clear technical English, and sustained research or engineering work demonstrated through projects, code, or papers.
Nice to have - Experience with egocentric video, geometry, multimodal temporal data, scalable annotation, or robotic manipulation. Publications are welcome but not required.
Benefits and work setup - Paid placement with close mentorship and a substantive project linked to potential infrastructure or model improvements. Scope, dates, and duration are agreed with the candidate.