Dec 11, 2025

Invited Vijay Janapa Reddi Petrut Bogdan jyik@g.harvard.edu steve.furber@manchester.ac.uk Eran Briman Pete Bernard raashid.ansari@silabs.com Ed Doran Rosina Haberl Chetan Singh Thakur EAIF Team Calendar Gideon Intrater klaus.knobloch@infineon.com friedemann.zenke@fmi.ch dhireesha.kudithipudi@utsa.edu n.beladel@student.tudelft.nl Venkat Rayudu Seung Hun SHIN/신승훈 jtapson@brainchip.com Frank Ghenassia Jason Eshraghian r.jolivet@maastrichtuniversity.nl andre.vanschaik@manchester.ac.uk

Attachments EDGE AI Neuromorphic WG EDGE AI Neuromorphic WG kick off meeting – 2025/09/29 08:56 CST – Notes by Gemini 

Summary

Petrut Bogdan presented on the well-received Ebrains meeting presentation on work-life balance and a career with randomness, and also noted that Jason Yeeke and another PhD student requested the resumption of NeuroBench work to target neuromorphic hardware. Participants including Rosina Haberl, Pete Bernard, Petrut Bogdan, and others discussed the planned neuromorphic conference day, aiming for mid-to-end of February, and suggested an 8:00 AM PST start time, with presentation lengths of 20-25 minutes plus Q&A. Jonathan Tapson, Klaus, and Steve Furber contributed to the discussion on content, suggesting topics like deploying systems “at scale” for business wins, exploring non-sexy applications, and including presentations on SpinCloud/SpiNNaker 2, and they identified SpinCloud’s hardware delivery and Jason Yeeke’s work on massive LLMs without matrix multiplication among the past year’s major neuromorphic achievements.

Details

Notes Length: Standard

• Ebrains Meeting and NeuroBench Update Petrut Bogdan reported that their presentation at the Ebrains meeting, which focused on work-life balance and embracing randomness in a career, was well-received by PhD students and echoed by a senior person. Petrut Bogdan also mentioned a request from Jason Yeeke and another PhD student to resume work on NeuroBench to modify the library to natively target neuromorphic hardware and include performance metrics. They proposed inviting them to future meetings in the new year to discuss required interfaces and the integration of other vendors like BrainChip.
• Neuromorphic Conference Day Planning Rosina Haberl raised the need to set a date for the planned neuromorphic conference day, suggesting Q1 of the next year. Petrut Bogdan clarified that this event would be similar to the generative AI working group’s conference day within the AGI foundation, and they are seeking input from the working group on its content. Rosina Haberl and Pete Bernard suggested an 8:00 AM PST start time for the live stream, running until 11:00 AM or 12:00 PM, noting that this generally works well across US and European time zones.
• Logistics and Scheduling for the Conference Day The ideal time slot for the conference day was discussed, with Pete Bernard noting that 8:00 AM PST, which is 4:00 PM to 8:00 PM UK time, seems to be a “sweet spot”. Rosina Haberl suggested 7:30 AM PST as a possible compromise. Petrut Bogdan proposed a tentative timeline around mid-to-end of February, specifically the week of the 16th to the 20th, which Pete Bernard confirmed is generally open.
• Proposed Topics for Conference Day: Commercial and Applied Neuromorphics Jonathan Tapson expressed willingness to participate and suggested discussing the impact of the “dramatic scaling of model sizes” at the edge on hardware and memory systems. While Petrut Bogdan thought an introductory talk on training and deployment might be beneficial for the Edgi Foundation community, Jonathan Tapson countered that their unique contribution would be discussing how to deploy systems “at scale” and secure business wins, offering an industry perspective not widely available from academics. Klaus indicated that their current work involves exploring non-sexy neuromorphic demo applications like battery management systems for microcontrollers, and they are seeking to understand the real-world gain of neuromorphic approaches in applications like cars.
• Inclusion of SpinCloud and Presentation Format Petrut Bogdan asked Steve Furber about the possibility of featuring SpinCloud or SpiNNaker 2, even though the latter is large-scale, arguing it counts as edge AI if the data is generated at the edge. Steve Furber suggested having someone from SpinCloud speak. Pete Bernard clarified that the conference live streams are open to anyone with valuable content and are not restricted to strategic partners, regardless of their startup status or funding. Rosina Haberl and Pete Bernard agreed that presentation lengths should be around 20 to 25 minutes plus Q&A, rather than the previous 15 minutes used when there were too many speakers.
• Major Neuromorphic Achievements in the Past Year When asked about the most important achievement in neuromorphics over the last year, Steve Furber cited SpinCloud actually delivering hardware. Jonathan Tapson highlighted several accomplishments, including Jason Yeeke’s work on “massive LLMs without matrix multiplication,” neuromorphic systems reliably flying on the International Space Station for over a year (specifically Greg Cohen’s DVS camera work), advances in training spiking neural models, and the contribution of neuromorphic engineers to Meta’s Ray-Ban Myoelectric hand interface. Klaus concurred with the significance of SpinCloud’s achievement.
• Future Outlook for Neuromorphics Jonathan Tapson expressed excitement about seeing new massive systems coming online, mentioning SpinCloud and Andre Fansky’s former group in Western Sydney. They also highlighted foundational Generative AI models based on spiking and event-based neural networks that could become competitive. Jonathan Tapson asserted that GenAI is coming to the edge and is rapidly increasing model sizes, stating that “the bitter lesson” of scale overwhelming other technical features is now affecting edge computing.