Purpose built for End to End High Performance Machine Learning​

Experience the power of HPC

Self Supervised Learning

1 %

Achieve accuracies with only 5-10% of labels; save cost, time and improve automation

Large scale Image Segmentation

0.001 b

Achieve better accuracies with Large DNN Models without any custom infrastructure work 

ScientificML to solve PDEs

1 x

Solve PDEs in real time, typically 40-100x faster than traditional HPC based simulations. 

RocketML is funded by

nsf logo

think beyond legacy. Welcome Transformation

Beyond supervision

Minimize cost of labeling by using state of the art self supervised and unsupervised techniques at scale

Autonomous Applications

Use data and experts to build autonomous applications that make predictions, detect anomalies

Autonomous insights

Use your vast amounts of unlabeled data to learn complex, highly nonlinear models with millions of parameters

Think Large Scale

AI at scale requires state-of-the-art infrastructure. Try our higher order based training for 4-10x speed ups

Solution Areas

Efficient Machine Learning infrastructure is a fundamental building block for building many applications at low TCO. It opens up opportunities for breakthrough discoveries and competitive advantage. Below is a very small sample of what is feasible on RocketML


With RocketML technology, insurance companies can move beyond their current level-2 automation. Insurance industry is information and document intensive. With RocketML unsupervised, large scale, continuous learning platform with built in integrations to popular insurance platforms, Insurance companies can modernize their applications quickly and easily


Ever increasing data sizes, velocity and uncertainties have made it imperative for researchers in Energy sector to embrace Machine Learning. RocketML offers highly optimized, end-to-end tuned “Machine augmented interpretations” of Seismic and other datasets. We take pride in demonstrating linear scaling to the limits of Amdahl’s law be it on GPU or CPU nodes 


The immune system is a uniquely challenging biological system to decipher. Our combinatorics-based data analysis has proven beneficial to immunologists leading to comprehensive, complete phenotype analysis of immune cells that was previously computational intractable. Immunologists can now address the most pressing needs, including new vaccines, improved cancer immunotherapies, and more

Frequently asked questions

RocketML enterprise version supports both Deep Learning and Traditional Machine Learning class of problems.

In addition, platform also supports data engineering tasks at scale. Typically data processing, transformation and cleaning tasks take up enormous amount of Data Scientists and Engineers time. RocketML makes performing these tasks at scale on GPU or CPU only clusters super easy. With built in functionality for deployment and integration of the models into Applications, RocketML truly the most Unified Science platform in the market.

RocketML is built to make Machine Learning on big data easy. RocketML supports on-the-fly compute cluster creation without preplanning, only when needed saving businesses enormous costs. Built ground up with HPC technologies, unlike other Distributed Machine learning tools, RocketML achieves strong scaling, a gold standard in HPC world, saving businesses greater than 50% in compute costs. 

Yes. We support GPU instances on Azure and AWS cloud. We can deploy our software into private on-premise data centers with GPU and CPU clusters. With a click of a button a Data Scientist or Engineer can start a GPU instance and use it an elastic, on-demand fashion saving thousands of dollars without compromising productivity

Unsupervised learning is the next frontier in machine learning technology innovation. Unlike supervised learning, unsupervised learning methods don’t require labeled data, thereby reducing data pre-processing related tasks by more than 50%. Our platform supports state of the art unsupervised machine learning techniques ranging from Self-Supervised to Semi-Supervised. 

While supervised learning has enabled many break throughs for both business productivity as well as decision making, Unsupervised learning is the best pathway to building Autonomous applications and insights.

Currently only the community edition is hosted. If an enterprises wants us to host our SaaS on cloud, we are happy to support the requirement at no extra cost. However, most of customers want us to deploy our SaaS into their private network be it on cloud or on-premise.

Yes. Our product is “highly secure” that not only protects Customers data but also their infrastructure. We have several security related innovation that makes our implementation highly desirable by regulated industries.

Yes. Absolutely!

Apart from superior scaling performance from RocketML, our customers also enjoy a completely hosted solution for Dask, Spark, Rapids, Tensorflow, PyTorch etc. at no extra cost!

It is truly the most versatile data science platform for both data engineering and data science work


Our clients and Partners say

"Profiling human immune system for cancer research or infectious diseases is a big data problem. We were shocked and delighted to experience 1000x speed up from RocketML, at lower cost without use of special hardwares like GPU"

Medical Researcher
"As a computational engine for handling data, @rocketml is superb. Learn how you too can explore data, build and tune models, and test them in production-like environments."

"To the best of my knowledge, RocketML is the first and the only Data Science and Machine Learning platform that brings HPC technologies to the Artificial Intelligence world. Because of these reasons, we picked RocketML over its much bigger alternatives"
Data Scientist
"We are pleased to welcome RocketML as one of our launch ISVs for the new AWS Marketplace for Machine Learning""Our customers want easy-to-use AI solutions running on Amazon SageMaker, and RocketML has provided a wide variety of algorithms customers can choose from directly from the Amazon SageMaker Console."