How HPE Ezmeral ML Ops addresses trends in data science and machine learning

[RichardHatheway]

Overview

In a recent report published by Gartner® (Data Science and Machine Learning Trends You Can’t Ignore, 27 Sept 2021)¹, they discussed key trends influencing the future of machine learning (ML) and data science (DS). This blog shows you how HPE Ezmeral ML Ops is already providing capabilities to allow you to stay ahead of these trends.

Trends and Gaps

Businesses today are beginning to use data science and machine learning (DSML) together. Because of this, organizations are starting to focus on deploying DSML platforms instead of just developing machine learning models from scratch. The reason is that over time these DSML platforms will evolve to provide greater access for different roles across the business. Of course, that means the expectations of what will be delivered by those platforms will also continue to increase.

The Gartner report indicates that “Data science is already mainstream, but in the world of ML, engineering is missing.” And also, “Companies familiar with the software development process expect quite a similarity with ML development process, but see many gaps on the engineering side. Filling in these gaps is what drives DSML trends today.”¹

The expectation is that ML should have similar processes to software development, but that isn’t the case. In most organizations today, machine learning projects don’t have the standard processes and rigor that are typically associated with software development. And while it may seem like a straightforward solution to use those DevOps tools and practices for the ML lifecycle, the reality is that ML workflows are very iterative in nature, so off-the-shelf software development tools and methodologies won’t work.

Data scientists still spend a significant amount of time and effort when moving projects from development to production. Model version control is still manual, making it hard to update models in production. Code sharing is manual, and data copied onto local storage leads to variability of results between environments. Additionally, the lack of standardization on tools and frameworks makes it tedious and time-consuming to ensure accuracy of predictions across all environments.

That’s why, as Gartner states: “We divided the trends into three groups, the endeavors which take organizations into the future. The group order is not sequential, but simultaneous. These endeavors underpin data science and ML platforms capabilities that enable organizations to:”¹

• Access to the platform for everyone — not just data scientists, but software engineers, business people, analysts, etc.
• Automate end-to-end DSML lifecycle
• Accelerate model development, deployment, and time to value

The bad news is that until HPE Ezmeral ML Ops came along, a DevOps-like solution didn’t really exist for machine learning projects.  

The good news is that HPE Ezmeral ML Ops exists today and will help you stay ahead of these trends.

Really? How?

Let me give you a quick overview.

The Solution

HPE Ezmeral ML Ops helps you automate your processes and accelerate your development and deployment by giving you an end-to-end data science solution that provides you with the prepackaged tools you need to operationalize machine learning workflows at every stage of the ML lifecycle, from pilot to production. This gives you the DevOps-like speed and agility your business is seeking along with the flexibility your business requires.

HPE Ezmeral ML Ops leverages the power of containers to create complex machine learning and deep learning stacks that include distributed TensorFlow, Apache Spark, H2O, and Python ML and DL toolkits, letting you spin-up distributed, scalable, machine learning and deep learning training environments in minutes rather than months. It also has a choice of programming languages and open-source tools to support even the most complex ML pipelines.

In addition, HPE Ezmeral ML Ops implements CI/CD processes for your ML projects with a model registry that stores models and versions, as well as storing models created using different tools/platforms. It also improves the reliability and reproducibility of machine learning projects via a shared source control repository (Github & BitBucket).

Finally, HPE Ezmeral ML Ops enables the deployment of models in production with secure, scalable, highly available endpoint deployment with out-of-the-box auto-scaling and load balancing. It enables the creation of custom application images with any combination of tools, library packages, and frameworks to suit your needs.

It also enables out-of-the-box application images to rapidly deploy containerized environments – sandbox, distributed training, or serving (inferencing) — with popular ML and DL tools, interfaces, and languages such as Python, R-Studio, TensorFlow, Spark, and more.

Summary

As you can see, these features allow data scientists to quickly and easily build and deploy machine learning models. Instead of managing infrastructure, they can focus on improving business outcomes.

HPE Ezmeral ML Ops lets data scientists quickly spin up environments with their choice of preferred data science tools, allowing them to explore a variety of data sources and simultaneously experiment with multiple machine learning/deep learning frameworks. They’re able to pick the best fit for the business problems that need to be addressed. And the good news is that all of this is ready to deploy in the enterprise!

HPE Ezmeral ML Ops integrates into existing environments and ties into enterprise security, code repositories, and image archives while providing multi-tenancy to meet the unique needs of every data science team in the organization.

To learn more about HPE Ezmeral ML Ops and how it can help your business, visit hpe.com/mlops or contact your local sales rep.

1 – Data Science and Machine Learning Trends You Can’t Ignore, published 27 September 2021, by Svetlana Sicular, Alexander Linden, Shubhangi Vashisth, Erick Brethenoux, Pieter den Hamer, Farhan Choudhary

If you’re a Gartner client, you can access the full report here.

Disclaimer:

GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally and is used herein with permission. All rights reserved.

Richard

HPE Ezmeral on LinkedIn | @HPE_Ezmeral on Twitter

HPE GreenLake on LinkedIn | @HPE_GreenLake on Twitter

@HPE_DevCom on Twitter