Postdoctoral Positions

POSITION 01

Learning for control of dexterous, contact-rich manipulation

Research scope

We seek a researcher with a strong background in reinforcement learning and control who wants to advance dexterous, contact-rich manipulation, with a particular interest in bimanual coordination. The scientific goal is to make learned manipulation behaviours adapt reliably to changing contact conditions, object properties and uncertainty during physical interaction.

Possible research directions include learning controllers that adapt to unfamiliar contact dynamics, combining tactile or force feedback with learned manipulation policies, and using pretrained robot policies within feedback control. Safety, robustness and recovery will be considered as part of developing and evaluating these methods.

The specific project will be shaped jointly around your expertise and interests, drawing on complementary work within PEARL on perception, representations and manipulation learning. You will have room to lead a focused research direction and develop further expertise in dexterous and bimanual systems.

Your research responsibilities

• Develop and lead a focused research project in learning for control of dexterous or contact-rich manipulation, contributing to the lab’s work on bimanual coordination.
• Design learning and control algorithms and evaluate how they adapt to changes in physical interaction, including their robustness and safety.
• Collaborate with colleagues on experimental evaluation in simulation and on physical robots, and contribute to scientific mentoring.
• Publish methodological contributions and communicate the assumptions, results and limitations of your work clearly.

Essential qualifications

• A PhD, completed or nearing completion, in robotics, control, machine learning, computer science, electrical engineering or a closely related field.
• Strong expertise in reinforcement learning, together with a solid grounding in feedback control and physical robotic systems.
• The ability to develop, implement and evaluate learning or control algorithms, supported by strong programming skills and evidence of independent research.
• The ability to reason about robustness, safety constraints and uncertainty in robotic execution.
• Familiarity with learned robot policies and an interest in connecting pretrained or foundation models to feedback control, together with clear scientific communication and a collaborative approach.

Publication record

Applicants should demonstrate substantial research contributions through peer-reviewed publications or accepted papers relevant to learning and control. We are interested in the questions you have addressed, the methods you have developed and your individual contribution to the work.

Publication venues and assessment

Relevant robotics conferences include the Conference on Robot Learning (CoRL), Robotics: Science and Systems (RSS), the IEEE International Conference on Robotics and Automation (ICRA) and the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Relevant journals include IEEE Transactions on Robotics, the International Journal of Robotics Research and IEEE Robotics and Automation Letters.

Methodological contributions in control venues, such as the IEEE Conference on Decision and Control (CDC) and IEEE Transactions on Automatic Control, or in major machine learning venues are also relevant when they connect clearly to learning and control for physical manipulation.

These venues are examples. Publications in comparable venues are welcome, and we will assess the quality and relevance of the research together with the applicant’s individual contribution.

Desirable experience

Experience in dexterous, contact-rich or bimanual manipulation is particularly welcome. Other useful strengths include tactile or force-based control, adaptive or model-based control, and adapting pretrained robot policies.

Candidates may bring different combinations of these strengths. Prior work across all of these topics is not expected; the project will build on your strongest methodological expertise and provide scope to develop complementary skills with the team.

Apply for this position ↓

POSITION 02

Structured and interactive embodied intelligence

Research scope

We seek a researcher to develop representations and learning methods through which robots acquire transferable physical knowledge from interaction. The research will connect structured representation learning, interactive perception and world models—models that predict how physical scenes change under actions—with the aim of improving generalization to unfamiliar objects, tasks and environments.

A central challenge is to identify what a robot needs to represent about objects, relations, contact and dynamics to support useful prediction and action. The position offers scope to investigate how geometric and physical structure can guide learning, how interaction can resolve uncertainty or reveal new concepts, and how these representations can connect to foundation models for perception, reasoning and action.

Your research responsibilities

• Develop an independent programme of research in structured representations, interactive world models or learning physical knowledge through embodied experience.
• Create methods that connect perception and action, using interaction to learn or refine representations and predictive models.
• Study which learned structure supports transfer across objects, tasks or environments, and design evaluations that identify why a method generalizes or fails.
• Collaborate with PEARL researchers on robotic evaluation, contribute to scientific mentoring and communicate methodological advances through publications and reproducible research.

Essential qualifications

• A PhD, completed or nearing completion, in machine learning, robotics, computer science or a closely related field.
• A strong background in machine learning and representation learning, with evidence of independent methodological research.
• Research experience in at least one relevant area, such as world models, structured or relational representations, geometric learning, interactive perception or physical reasoning.
• An ability to formulate precise research questions, examine modelling assumptions and design informative experiments.
• Strong programming skills, clear scientific communication and an interest in connecting fundamental learning methods with embodied systems.

Publication record

Applicants should have a strong record of peer-reviewed publications or accepted papers demonstrating substantial scientific contributions relevant to the position.

Publication venues and assessment

Relevant machine learning conferences include Neural Information Processing Systems (NeurIPS), the International Conference on Machine Learning (ICML) and the International Conference on Learning Representations (ICLR). Work at the Conference on Robot Learning (CoRL) and Robotics: Science and Systems (RSS) is particularly relevant to the position’s connection with embodied systems.

For contributions in perception and representation learning, relevant venues also include the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) and the IEEE/CVF International Conference on Computer Vision (ICCV), as well as comparable peer-reviewed journals.

These venues are examples. Publications in comparable venues are welcome, and we will assess the quality and relevance of the research together with the applicant’s individual contribution.

Desirable experience

Relevant experience may include object-centric learning, geometric or physics-informed models, learning dynamics, active exploration, multimodal learning, causal representations, foundation models or real-robot experimentation. We welcome applicants with a strong machine learning background who want to address fundamental questions through interaction with the physical world.

Apply for this position ↓

APPLICATION

Apply to PEARL

Choose either research track or indicate an interest in both. Please prepare:

• Motivation: up to 2,000 characters explaining your interests and fit to the selected position.
• One PDF: your CV, a brief research statement and publication list. Explain the questions you would like to pursue at PEARL and how your previous work prepares you to address them.
• Referees: two names and professional email addresses, with an optional third. Obtain their agreement before listing them.
• Background and availability: your current affiliation, PhD completion or expected completion date, and earliest possible start date.

You may also add links to up to three selected publications with a short explanation of your contribution, and relevant code, project or personal website links. Ensure that the selection committee can access the PDF.

Questions? contact@pearl-lab.com

This call is for two postdoctoral positions. We are not currently recruiting PhD students.