Research and development in intelligence, machine learning, and neuromorphic engineering by Jens Egholm Pedersen.

Research Focus

I develop spatiotemporal computation frameworks for brain-inspired systems, bridging neuroscience and artificial intelligence. My work focuses on creating practical tools and theoretical foundations for neuromorphic computing.

Key Projects

Read more about my research

Open Source Contributions

I maintain several core libraries for the neuromorphic computing community:

NIR

Standard for bio-inspired computing

GitHub stars Image Image

Norse

Spiking neural network library

GitHub stars Image Image

AEStream

Streams sparse event data

GitHub stars Image Image

Faery

Process event camera data

GitHub stars Image
View all projects Sponsor my work

Speaking & Media

I work to increase visibility of neuromorphic computing in AI and broader scientific communities through conference presentations, interviews, and public engagement.

Watch talks and media appearances

Recent Research

image from Spiking neural receptive fields
- 65 words - 1min read

Spiking neural receptive fields

I was invited to give student talk on at the Open Neuromorphic student talk about spiking neural receptive fields. My talk expands on our work on spatio-temporal covariance for leaky integrators and leaky integrate-and-fire neurons published at Nature Communications in 2025. I motivated the problem about neuromorphic computing, talked about the necessity for covariance guarantees, and discussed outlook and next steps. Slides are available here.
Read the full entry here
image from Covariant spatio-temporal receptive fields for spiking neural networks
- 1124 words - 6min read

Covariant spatio-temporal receptive fields for spiking neural networks

Neuromorphic computing exploits the laws of physics to perform computations, similar to the human brain. If we can “lower” the computation into physics, we achieve extreme energy gains, up to 27-35 orders of magnitude. So, why aren’t we doing that? Presently, we lack theories to guide efficient implementations. We can build the circuits, but we don’t know how to combine them to achieve what we want. Current neuromorphic models cannot compete with deep learning.
Read the full entry here

More

Where Newton and Turing meet: neuromorphic computing through mixed-signal systems

Telluride: Dynamics for control systems

Science is open software

Tonik: Neuromorphics on the Faroe Island

All Research