James Heald

I develop algorithms for learning dexterous manipulation with multi-fingered hands, combining reinforcement learning and probabilistic machine learning. Most recently, this led to a NeurIPS 2024 competition-winning method that culminated in an ICML 2026 Spotlight paper (top 2.2% of submissions).

I am a Research Fellow at the Gatsby Unit, UCL working with Maneesh Sahani. I previously worked as a postdoctoral researcher at the Zuckerman Institute, Columbia University with Daniel Wolpert. I received a PhD in Engineering from the University of Cambridge, where I trained in the Computational and Biological Learning Lab. Prior to my PhD, I worked as a medical doctor in the NHS, and before that, I received an MSc in Cognitive and Computational Neuroscience from The University of Sheffield and an MBChB in Medicine from The University of Manchester.

Email  /  CV  /  Google Scholar  /  GitHub  /  LinkedIn  /  Twitter

I am an active contributor to MyoSuite, participating in biweekly development meetings. Working collaboratively and independently, I have:

• Ported environments from MuJoCo to MuJoCo XLA, enabling massively parallel, GPU-accelerated simulations.
• Designed new musculoskeletal models and tasks to expand the toolkit of the wider research community.
• Built an API to allow users to easily modify existing MuJoCo models for their own research needs.
• Created an international advocacy website to widen participation in the NeurIPS 2024 MyoChallenge competition.

Outstanding research achievements are highlighted below.

MyoChallenge 2024: A New Benchmark for Physiological Dexterity and Agility in Bionic Humans
NeurIPS 2024 Competition Track Program
James Heald, Kai Biegun, Samo Hromadka, Maneesh Sahani
First Place
Paper

Our team, Muscle Heads, came first place in the Manipulation Track of the NeurIPS 2024 MyoChallenge competition, winning a $2,000 prize. The challenge involved coordinating a biologically-realistic human arm with 63 muscles and a robotic prosthetic arm with 17 DoFs to transfer objects between pillars. To control the musculoskeletal arm, I developed a novel algorithm to learn low-dimensional control manifolds in high-dimensional action spaces. This allowed us to efficiently acquire dexterous reaching and grasping skills. To control the robotic prosthetic arm, we combined a learned policy with inverse kinematic control, enabling smooth hand-offs and object relocation. Curriculum learning and reward shaping were used to solve the task in stages (reach, grasp, handover, move object to goal) .

We develop an information-theoretic framework for discovering low-dimensional action manifolds in tendon-driven systems. Across both musculoskeletal and robotic hand models, policies trained on these manifolds achieve substantially greater dexterity, sample efficiency and generalization.

We review a computational framework for repertoire learning that provides a unifying account of phenomena across numerous domains, including motor learning, economic decision making, episodic memory and conditioning.

We develop a theory of motor learning based on the principle of contextual inference. Our theory reveals that adaptation can arise by both creating and updating memories and changing how existing memories are differentially expressed.

website template