As Natural Language Processing (NLP) models evolve to become ever bigger, GPU performance and capability degrades at an exponential rate, leaving organizations across a range of industries in need of higher quality language processing, but increasingly constrained by today’s solutions.

Throughout their lifecycles, modern industrial NLP models follow a cadence. They start from one-time task-agnostic pre-training and then go through task-specific training on quickly changing user data. These periodically updated models are eventually deployed to serve massive online inference requests from applications.