Darius Pouzlay

'I'd like to buy the world a chip and keep them company...' Run Akida anywhere. Run Akida everywhere. Not just a slogan anymore.

I want to extend a heartfelt thanks to my great friend and research colleague Dr. Steve Harbour, PhD, Technical Fellow at BrainChip, who was able to attend IEEE NAECON 2026 last week when I wasn't. He successfully presented our three accepted papers which will be published by IEEE in October. We've been busy developing work around neuromorphic computing, witnessing the expansion of BrainChip's Akida into larger AI-system architectures. Steve spoke to a packed house and had lots of dynamic conversations about how neuromorphic processors are moving into enterprise compute where it stands to help lower the power footprint significantly, increase the efficiency and utilization of current compute grids in conjunction with IBM Spectrum Symphony and GPFS, and help institutions and companies expand intelligently. Our papers explored neuromorphic orchestration for enterprise LLM infrastructure, real-time ISR and multi-sensor fusion, alongside Steve's broader work on decision-optimized heterogeneous AI architectures. Simply put, the conversation has moved beyond "edge" to consider what becomes possible when event-driven neuromorphic inference is represented as a first-class compute tier within a much larger heterogeneous AI architecture in the cloud, in data centers, and on-prem. That's where I believe this gets interesting as my demo record makes clear. The photographs Steve shared with me show some of the technical discussions surrounding the work including engagement with researchers from RIT and the University of Dayton as well as him speaking at the event. I've gone ahead and attached more detailed images of the posters I created for you to review. We are now deep into this continuum and we will continue to see things progress as we go: research → architecture → applications → validation → adoption Akida's future opportunity is considerably larger than what is typically called edge AI when coupled with a heterogeneous compute ontology. The work we're exploring around LLM infrastructure, distributed inference, ISR, sensor fusion, and heterogeneous compute orchestration shows why. Congratulations to Steve and everyone involved with IEEE NAECON 2026. I'm looking forward to continuing the research and seeing this break out toward a new computing future.

Finally had time this morning to switch out and place four AKD1000's. Quick shot of them running like clockwork now on a $27 NVME expansion board on an AMD EPYC 7702P system. So beautiful...the serpentine cables around them go to 3090s. Heterogeneous compute ontology for the win.