OctoML Inc. today announced that it has closed $15 million in fresh funding, a mere six months after last raising capital, to continue work on its automated artificial intelligence optimization platform.

Seattle-based OctoML was founded last year by a group of University of Washington computer scientists. The team, led by Chief Executive Officer Luis Ceze (pictured, second from the left), has developed an open-source framework called Apache TVM that helps developers optimize their AI models for the chips on which they are destined to run.

Apache TVM improves the performance of neural networks by a factor of 10 or more in some cases by fine-tuning them for a processor’s specific architecture. It supports a range of chips, from graphics cards to more specialized AI accelerators. The framework has been adopted by developers at companies such as Facebook Inc., Microsoft Corp. and Amazon.com Inc., which uses it to optimize parts of Alexa.

OctoML is looking to commercialize Apache TVM’s success with a paid cloud-based solution called Octomizer. It’s essentially a more user-friendly version of the framework, allowing developers to have a model optimized by simply uploading their code and specifying the chips for which they want to optimize it.

The startup’s pitch to enterprises is that Octomizer can make AI development more efficient. Model optimization normally is a manual task that requires developers to fine-tune their code painstakingly for “vendor-specific software stacks and gaps in operator coverage caused by fast model evolution,” Ceze wrote in a blog post today. Automating the process “reduces the time taken to put models into actual use, reduces edge resource” and cuts cloud bills, he added.

OctoML is currently piloting Octomizer with early customers. The startup plans to use the new funding to keep building out the platform and expand its workforce, which has grown from six employees to 20 in recent months.

OctoML raised the investment from Amplify Partners Madrona Venture Group, the same two venture capital firms that led its previous funding round six months ago. 

Photo: OctoML