Commercializing a programmable, fairly general purpose, computer design with 50x better speed/watt than GPUs. Suited to deep learning (neural nets), vision, speech, pattern recognition, but also optimization, simulation, and other traditional domains. Scales from embedded systems (100,000 core mobiles, cars, robots) to data centers (hundreds of millions of cores per rack).Singular's approximating computing inventions are widely used in modern AI hardware, and its patents have been licensed by large companies including Google.

Exploring applications of the Singular Computing machine to computer vision and image processing.

Spinoff from my Carnegie Mellon research group, funded by Fujitsu and US Commerce Dept, to develop and commercialize interactive character and interactive drama technology, along with content. Praised by NYTimes and others, deployed by McDonald's and others, but tech was too expensive to succeed in the commercial world during the Internet bubble crash ~2001.

Research Professor in Computer Science Department, and Fellow in College of Fine Arts. Created and led Oz Project, a fairly well regarded effort to develop powerful interactive story and character technology. Got promoted to Full Prof rank and promptly left, when Zoesis opportunity arose. You only go 'round once in life...

Co-founder of PRL project (name came from my dissertation), which was led by Robert Constable. Helped develop NuPRL logic and led first implementation. This remains a well regarded system for doing semi-automated mathematics, in a very rich proof framework based on constructive type theory.