Michael James

Michael James possesses a deeply specialized background in high-performance computing and advanced software engineering. Having previously partnered with the founding team as a lead engineer at SeaMicro and AMD, his massive background in optimizing the communication pathways between hardware and software was absolutely essential in developing the custom compilers and software stack required to actually program the unprecedented Cerebras hardware.

Michael James is a co-founder and Chief Software Architect at Cerebras Systems. He directs the Cerebras software compiler stack, developing automated graph compilation tools that translate PyTorch and TensorFlow neural networks into physical core topologies across the Wafer-Scale Engine with zero manual parallel programming required from machine learning developers. At Cerebras, James directs compiler software engineering, developing the automated graph compilation algorithms that map standard PyTorch and TensorFlow neural networks directly across the 900,000 cores of the Wafer-Scale Engine. His compiler architecture abstracts away the agonizing complexity of distributed parallel programming, allowing researchers to treat an entire wafer supercomputer as a single unified processor.