OctoML provides an automated deployment machine learning software platform to businesses, designed to optimize and speed up the performance of AI solutions at the network edge. OctoML was founded in 2019 by Luis Ceze, Jared Roesch, Jason Knight, and Thierry Moreau.
At TVMCon 2021, OctoML announced the release of its new machine learning automated deployment platform, an edge AI solution for performance optimization. It is designed for businesses to automate and scale deploying machine learning models to a range of cloud services including AWS, Microsoft Azure, and Google Cloud Platform, as well as hardware infrastructure like Nvidia GPUs, Intel CPUs, and AMD CPUs.
OctoML's platform is built on the open-source compiler stack Apache TVM. The platform pairs with individual optimization models and automatically optimizes them to fit with the selected hardware without having to manually restructure it. It also gives access to comprehensive benchmarking, in which the newly optimized model can be compared against the original or similar public models and various CPU and GPU instance types. Its seamless automated deployment feature works with just a few command lines of code, saving businesses time otherwise spent on manual optimizations and performance testing.
OctoML's platform can be used to run natural language processing (NLP) models to assist with virtual assistant development and customer sentiment analysis.
The platform helps provide solutions for various computer vision field applications like intelligent cameras, self-driving cars, medical imaging, and video conferencing backgrounds.