Jonathan Kao
University of California-Los Angeles, CA·Biomedical Engineering·Early-career
How do networks of neurons in the brain turn intentions into decisions and movements, and can that knowledge help people with paralysis move again? This work connects these questions using artificial intelligence.
- Fields
Biomedical Engineering
Neuroscience
Computer Science
- Topics
brain-machine interface
brain-computer interface
neural decoding
- Grants
- 3 known grants, through Aug 2030
What they work on
- Train new recurrent neural network architectures on the decision-making and movement tasks animals perform, constrained by anatomy and neural recordings, then analyze them to propose mechanisms of cortical computation.
- Test hypotheses generated by brain-like recurrent neural network models in collaboration with experimental labs and build the insights into better brain-computer interface decoders.
- Build recurrent neural network models that generate synthetic motor cortex activity for multi-joint reaching, trained on human clinical trial data, as the core of an open-source brain-machine interface simulator.
- Model how human users adapt to brain-machine interface decoders using deep imitation and reinforcement learning, so the simulator can predict decoder performance in minutes instead of months.
- Train AI copilots that combine computer vision with fused non-invasive signals of user intent to infer goals and help people with paralysis perform complex everyday tasks.
- Develop methods for a user and an AI copilot to adapt to each other during shared control, testing whether this improves system performance and robustness.
More about the research
Projects range from building brain-like computer models, to simulating brain-machine interfaces, to designing AI "copilots" that help paralyzed users complete everyday tasks.
Why it matters
People with paralysis or ALS have few devices that meaningfully improve daily life. This work could lead to low-risk, non-invasive assistive devices that help them perform everyday tasks. An open-source BMI simulator could speed up decoder design and let many more researchers contribute. Models of brain computation could also inform diagnosis and treatment of cognitive and motor disorders.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Restoring complex movement and locomotion after paralysis through collaborative copilots
NIH · Sep 2025 – Aug 2030
An Open Source Simulator for Multi Degree-Of-Freedom Brain-Machine Interfaces
NIH · May 2021 – Apr 2027
CAREER: Elucidating principles of cortical computation with recurrent neural networks
NSF · Oct 2020 – Sep 2026
Closest labs
Lee Miller
Northwestern University, IL·Biomedical Engineering
Build decoders that predict intended muscle activity (EMG) from primary motor cortex spiking recorded in humans with high-level spinal cord injury, trained with EMG, hand posture and force data from able-bodied people doing the same actions.
Karunesh Ganguly
University of California-San Francisco, CA·Neuroscience
Compare two neural decoding methods for their ability to give paralyzed patients stable, plug-and-play control of a robotic arm and hand with ECoG for at least 8 weeks without recalibration.
Sergey Stavisky
University of California-Davis, CA·Neuroscience·Early-career
Record hundreds of individual neurons with implanted microelectrode arrays in the ventral precentral gyrus and superior temporal gyrus to learn which features of speech these neurons encode and when during speaking.
The school
University of California-Los Angeles: 84 labs on PI Pages, $1.9B a year on research, 803 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.