An AI research lab, built where the problems are.

David has spent eight years architecting distributed systems serving more than 5 million monthly active users, across national infrastructure, regulated finance and public-sector platforms. At the lab he researches how models actually get deployed: inference that holds under real load, systems that keep working offline, and what it takes to run them on hardware an institution already owns. He builds the engineering teams that keep them running.

Daniel is a researcher first. His work spans the width of applied AI: computer vision and object detection, optical character recognition, face and gesture recognition, speech-to-text, natural language processing and recommender systems, built end to end from data preparation through deployment and evaluation. The through-line is efficiency, driving deep networks and CNNs down to run inside hardware budgets that would normally rule them out, which is the exact problem this lab exists to solve. He sets the lab's research agenda and the standard it reports to: experiments run to a protocol, results measured on the hardware institutions actually own, and findings published including the ones that did not work.