Fractile

We are building the hardware that will remove every bottleneck to the fastest possible inference of the largest transformer networks thanks to in-memory computing.