About the Fellowship
The Postdoctoral Fellow position on the MARA (Modeling, Abstraction, and Reasoning Agents) project offers an immersive research environment to advance the frontiers of active world-model learning applied to robotics. Working under the mentorship of Tom Silver and alongside the MARA team, you will develop novel algorithms that enable robots to actively explore and build models of their environment through principled experimentation.
About Basis
Basis is a nonprofit applied AI research organization with two mutually reinforcing goals.
The first is to understand and build intelligence. This entails establishing the mathematical principles of reasoning, learning, decision-making, understanding, and explaining, and constructing software that embodies these principles.
The second is to advance society’s ability to solve intractable problems. This involves expanding the scale, complexity, and breadth of problems we can solve today and, more importantly, accelerating our ability to solve problems in the future.
To achieve these goals, we are building both a new technological foundation inspired by human reasoning, and a new type of collaborative organization that prioritizes human value.
About Tom Silver’s Group
Tom Silver is an Assistant Professor of Electrical and Computer Engineering at Princeton University. His research focuses on learning and planning for long-horizon robotic tasks, with a particular emphasis on neuro-symbolic methods, program synthesis, and active learning.
Research Focus
We are looking for exceptional early-career researchers who combine theoretical depth with practical ambition. The ideal candidate has completed (or is near completing) a PhD in a relevant area, has demonstrated research excellence through publications, and is excited about the unique challenges of embodied intelligence. You will co-develop a focused, jointly scoped project with Tom Silver’s group that advances MARA’s broader mission of creating agents that learn like scientists.
This fellowship offers the unique opportunity to validate theoretical advances on real robotic systems, working closely with our hardware team to ensure algorithms translate from simulation to reality. You will be part of shaping how robots can efficiently learn about their world through active experimentation.
The fellowship is time-limited, typically spanning 1 to 2 years, with potential transition to a permanent Research Scientist position based on mutual interest and performance.
Who we’re looking for
- Hold (or be near completion of) a PhD in technical areas such as robotics, machine learning, computer science, cognitive science, or related fields.
- Have demonstrated research excellence through publications at top venues (NeurIPS, ICML, RSS, CoRL, ICRA) or significant contributions to the field.
- Possess deep knowledge of relevant research areas including active learning, intrinsic motivation, exploration in RL, experimental design, or information-theoretic approaches to learning.
- Be capable of independent research while valuing collaboration. You can drive your own research agenda while contributing to team objectives.
- Have strong theoretical foundations combined with implementation skills to validate ideas experimentally.
- Be excited about bridging theory and practice by testing algorithms on physical robot systems and addressing the challenges of real-world deployment.
In addition, the following would be an advantage:
- Prior experience with robotic systems and hardware experiments.
- Publications specifically in active learning, exploration, or experimental design.
- Contributions to open-source robotics or ML projects.
- Demonstrated ability to bridge multiple research areas.
- Experience with probabilistic modeling and inference.
- Strong communication skills for both technical and general audiences.
Core Responsibilities
- Design and develop novel algorithms for active world-model learning, focusing on how agents can efficiently explore environments and learn causal structure through experimentation.
- Conduct independent research within the MARA framework, pursuing questions at the intersection of active learning, robotics, and scientific reasoning.
- Validate algorithms on robot hardware by working with the hardware team to implement and test approaches on physical systems.
- Collaborate with the MARA team to integrate active learning methods with world modeling, planning, and control components.
- Publish and present research at top conferences.
- Mentor junior researchers and contribute to the intellectual culture of Basis through seminars, reading groups, and collaborative projects.
- Contribute to open-source software that enables the broader research community to build on your work.
- Help define future research directions for MARA and potentially develop independent research proposals.
- Collaborate with MARA robotics teams and with teams in Tom Silver’s lab to validate ideas across both environments.
Role Details
- Position Type: Time-limited fellowship (1-2 years)
- In-person Policy: We are in the office four days a week. Be prepared to attend multi-day Basis-wide in-person events.
- Location: New York City.
- Collaboration sites: Regular collaboration at Princeton University; flexible time partitioning between Basis (NY) and Princeton based on project needs.
- Mentor: Tom Silver (primary), with additional mentorship from MARA team leads.
- Salary: Competitive with leading postdoctoral fellowships.
- Research funds: Access to an annual allocation for research expenses, conference travel, and compute.
- Start date: Flexible based on candidate availability.
- Career path: Potential transition to permanent Research Scientist position.
Application Materials
- CV including publication list
- Research statement (2-3 pages) outlining past work and proposed research directions
- Three letters of recommendation
- One representative publication or technical report
Privacy Notice
By submitting your application, you grant Basis permission to use your materials for both hiring evaluation and recruitment-related research and development purposes. Your information may be processed in different countries, including the US. You retain copyright while providing Basis a license to use these materials for the stated purposes. Read our full Global Data Privacy Notice here.