Deep Learning (DL) technology has matured to a point where it achieves remarkable accuracy on many human-related tasks, however, at an insurmountable computational cost: It is predicted that by 2026, the energy cost of training a state-of-art vision or NLP model, would be equal to the monthly carbon-deoxide emission of New-York City (!), and therefore scalable training and inference of deep neural networks (DNNs) is one of the major challenges of modern AI (and humanity).

I am collaborating with Deci.AI (https://deci.ai) to accelerate DNNs by integrating novel data structures (which we term "Neural Data Structures") into neural nets, in order to dramatically accelerate matrix-multiplication operations in DNNs with negligible accuracy loss.
Our research group at Deci includes world-class CS researchers and engineers (from HUJI, Weizmann and Technion), and is building proprietary algorithmic technology. If you are an enthusiastic developer or researcher with experience in Algorithms & AI, please reach out to me directly!