Virtualization Technology News and Information
Verta 2021 Predictions: MLOps Trends and Impact on Enterprise Data Science Teams

vmblog 2021 prediction series 

Industry executives and experts share their predictions for 2021.  Read them in this 13th annual series exclusive.

MLOps Trends and Impact on Enterprise Data Science Teams

By Manasi Vartak, Founder & CEO at Verta

The term MLOps or machine learning operations emerged in 2020 and suddenly became widely used in relation to AI and machine learning. With more and more enterprises starting to implement AI technologies in their organizations increases the need for tools and processes to efficiently develop and deploy production-ready ML models, which combine development-oriented activities (Dev) with IT operations (Ops).

MLOps is a set of methods, which facilitates the application of DevOps to machine learning and enables data scientists and ML engineers to robustly version ML models, collaborate and share ML knowledge, and when models are ready for graduation, to deploy and monitor models in production environments. Fortunately, great MLOps tools help enterprise data science teams to take models from research to production while simplifying the ML development process.

How will this trend develop and what can organizations expect from MLOps in the upcoming months? Here's some predictions for 2021.

1. ML Models becomes foundational to many sectors

Models are the new code. They will routinely implement functionality that previously needed to be hand-written (e.g., database indexing) and serve as the foundations of key product features (e.g., searching video content). In fact, ML models already generate millions of dollars in revenue and are likely to become foundational to many sectors including finance, security, and healthcare.

2. Data Scientists move closer to DevOps

The goal of ML is to help products make decisions on the spot. For example, which messages to show from a search query. These complex applications focus on actively improving the business instead of just providing insights and require the combination of different skill sets within organizations.

With MLOps in practice, data scientists and DevOps teams move closer together and bring some major advancements to model development:

  • Data scientists increasingly own their work end-to-end. Instead of sending their models to a software developer and data engineer to build the machinery on a case by case basis, they can easily integrate it into MLOps platforms.
  • Data scientists are becoming aware of the processes required to satisfy operations and can increasingly avoid that a work is dropped or re-done.
  • ML deployment will work seamlessly with DevOps pipelines, giving Operations folks a peace of mind.

3. ML models become scalable

In contrast to the frequently brittle ML pipelines that are the result of one-off model deployment efforts, ML models become scalable and production-ready from Day 1: models operationalized via an MLOps platforms are robust, ready for high performance, and can be tightly integrated into the rest of the software or IT infrastructure from Day-1. The models work for you for Production Model #1 and continues to scale and grow all the way to Production Model #10,000.

4. Model assets are no longer hiding in the dark 

Organizations with hundreds of models (e.g. banks, insurance companies) face a unique type of challenge. Arising from the heterogeneity in ML workflows and the siloed nature of large enterprises, there has so far been no way to keep track of all model assets across the organization. For example, it was near-impossible to answer questions such as what models exist across the organization, where a given model is being used, who is using that model, and whether that model is performing as expected. MLOps tools give organizations visibility to their models to avoid duplicated effort in data science, and leave them no longer open to liabilities arising from unintended model use.

Seeing the great impact of MLOps on the development lifecycle of ML models across many disciplines, we can definitely say that this trend is here to stay and MLOps will gain even more importance in 2021. Thanks to ready-to-use MLOps tools, development and deployment of ML models will become less complex and hopefully much more ML products will successfully make it to production.


About the Author

Manasi Vartak, Founder & CEO at Verta

Manasi Vartak 

Manasi Vartak is the founder and CEO of Verta, an MIT-spinoff building software to enable high-velocity machine learning. The Verta platform enables data scientists and ML engineers to robustly version ML models, collaborate and share ML knowledge, and when models are ready for graduation, to deploy and monitor models in production environments. Verta grew out of Manasi's Ph.D. work at MIT on ModelDB, the first open-source model management system deployed at Fortune-500 companies. Manasi previously worked on deep learning for content recommendation as part of the feed-ranking team at Twitter and dynamic ad-targeting at Google.

Published Friday, October 30, 2020 7:37 AM by David Marshall
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
<October 2020>