Anyone who has been following his projects knows that 2.5 years is too little a time for the problems he set out to solve and the projects he was leading (esp.

Anyone who has been following his projects knows that 2.5 years is too little a time for the problems he set out to solve and the projects he was leading (esp. with TensorFlow, MLIR, TF runtime, TPUs). So this is really a very early, abrupt and surprising exit, and several people both inside/outside know that he ran into political issues/struggles with high-level folks. Fortunately, one of the projects (MLIR) is already part of LLVM and will remain unaffected, but looks like S4TF may just die out irrespective of what's being said. Otherwise, we may not hear much on the details given his "seemingly" reasonable length stint at Google.