New Deep Learning Startup, OctoML, Launches to Automate Deployment of Secure and Efficient Deep Learning to Cloud and Edge
Apache TVM is an automated deep learning model optimization and compilation stack that powers efficient model deployment in major technology companies like Amazon, Facebook, Microsoft, Xilinx, and Qualcomm.
OctoML today announced $3.9 million in seed financing led by Madrona Venture Group with participation from Amplify Partners, to bring the ability to deploy deep learning models on every hardware configuration to companies across the globe. OctoML is leveraging the power and traction of Apache TVM, an open source project originated by the founding team, to enable companies of every size to harness the power of deep learning without the expensive heavy lifting of tuning and securing models to each hardware configuration that a customer might need.
Apache TVM is an automated deep learning model optimization and compilation stack that powers efficient model deployment in major technology companies like Amazon, Facebook, Microsoft, Xilinx, and Qualcomm. It is backed by a thriving community of more 270 contributors worldwide, including people from major tech companies and academic institutions. The mission of the project is to enable data engineers to easily optimize and deploy models across a broad set of hardware in a portable manner.
“It has been awesome to see the community take off around Apache TVM. It is clear that researchers and large tech companies see both the utility and value it can offer in pushing the state of the art in efficiency machine learning in research and production settings,” said Luis Ceze, co-founder and CEO. “The promise of Apache TVM is so high that hardware companies are actively looking to optimize for it and companies with lean teams need solutions to not only deploy these models but keep them up and running. That is the opportunity that OctoML is pursuing. OctoML’s automated model optimization technology leveraging TVM, leads to lower engineering and operating costs for customers, and lowers the risk of dependence on specific platforms.”
OctoML will offer a managed service for companies looking to securely deploy in multi-cloud and edge environments and ensure that the models stay up and running at peak efficiency.
“Intelligent applications are changing the landscape of software – and in fact blending again the roles of software and hardware. The work that the OctoML team has done to build this technology into a powerhouse in the open source community is outstanding and we are excited to back this team,” said Matt McIlwain, Managing Director, Madrona Venture Group.
Why Apache TVM?
Deploying deep learning models efficiently is challenging because there is a fast-growing set of model architectures and operations that need to be mapped to a fast growing set of hardware targets such as CPUs, GPUs, FPGAs and a multitude of specialized hardware accelerators. Optimizing models for hardware targets today is typically done manually, a laborious but profitable process that can yield orders of magnitude gains in terms of better performance and energy efficiency. Apache TVM’s magic to address this problem is using machine learning to automate ML code generation and optimization.
“Apache TVM’s machine-learning based approach to optimizing machine learning systems fundamentally enables targeting a constantly changing and expanding set of hardware targets such as data centers, cars, phones, health devices and embedded systems with much less engineering,” said Tianqi Chen, co-founder and CTO of OctoML.
“OctoML is taking several steps further in automation and is ramping up work with hardware vendors to optimize the TVM stack for their current and upcoming chips,” said Jason Knight, Chief Product Officer and co-founder of OctoML, and former Head of Software Product - AI Products Group at Intel.
The company, a spinout of the University of Washington Allen School for Computer Science and Engineering, is being led by the original creators of the TVM technology. The founding team Luis Ceze, Jason Knight, Tianqi Chen, Thierry Moreau and Jared Roesch came together around their work on TVM after they recognized a need in the industry for a solution to enable easy, efficient and secure deployment of deep learning models, and that Apache TVM can serve as a foundation for that. The company also counts as an advisor Carlos Guestrin, a recognized machine learning leader at the Allen School and in his role as director of AI and Machine Learning at Apple.
About OctoML
OctoML (https://octoml.ai/) offers easy, efficient and secure deployments of machine learning and deep learning models as a managed service. The company is based in Seattle.
About Madrona Venture Group
Madrona invests predominantly in seed and Series A rounds across the information technology spectrum, with a focus on infrastructure, intelligent applications and ML, ecommerce, and consumer services. Madrona manages nearly $1.7 billion and was an early investor in companies such as Amazon, Apptio, Rover.com, and Redfin. www.madrona.com
View source version on businesswire.com: https://www.businesswire.com/news/home/20191023005226/en/