We were researching adversarial training on imbalanced multi-modal classes in 2010. We built hardware-software codesigned AI infra before that was cool (spoiler alert: it wasn’t unicorn seed rounds, but it had lots of FPGAs and CPUs and some GPUs). Some of what we built went big, some failed. A lot of our research hit a wall or remains a what-if.
Research is not linear. Neither are brilliant minds. That’s just not how it works when inventing the algos, middleware, ledgers & databases, system and software architectures, hardware of the future. We believe that the best way to unleash your brain is to bring you the right resources in large enough quantity while you retain the flexibility and freedom from inside your lab.
What problems you see worth solving and how you think through solving them is all that matters to us.