Petuum builds software that automates data preparation and model selection and virtualizes hardware so developers can optimize machine-learning models for specific hardware constraints. The platform aims to make AI development accessible to novices while providing standardized, composable building blocks for experts. The company helps users avoid low-level management of distributed GPU clusters and simplifies use of frameworks like TensorFlow and Caffe. Founded last year by Dr. Eric Xing, Dr. Qirong Ho and Dr. Ning Li, Petuum is initially targeting healthcare and fintech customers and is working with beta testers across industries. The company currently has about 70 employees and says it will expand product, sales, and marketing simultaneously. The article notes market skepticism about horizontal AI platforms but highlights Petuum's significant capitalization relative to peers. Petuum builds a general-purpose AI/ML development platform that supports deep learning, predictive analytics, knowledge extraction, content summarization, ensemble methods and a wide range of applications such as NLP, image/video understanding, and anomaly detection. The platform is designed to integrate and process diverse data sources including social media, consumer profiles, EHRs, IoT sensor logs, financial transaction logs, and manufacturing machine logs. Petuum’s architecture emphasizes portability — compiling code into a universal format and mounting executables across hardware from data-center scale systems to mobile devices — and uses a parameter server, managed communication, adaptive scheduling, load balancing, and elastic resource management. The company says its multi-tenant platform is 10 to 100 times more efficient than alternative solutions and is significantly more efficient than Hadoop and Spark for running AI/ML applications. Petuum positions itself as both an AI/ML development platform and an AI/ML computing management and operating system for enterprise deployments. The company has assembled an advisory board of academic and industry experts and announced a $15 million Series A to support its growth.

Total raised

$108M

Funding rounds

2

Latest round

Series B

Latest activity

Oct 2017