VK Cloud ML Platform

The main articles are:

• SaaS - History. Philosophy. Development drivers
• Data Quality Management

VK Cloud (formerly VK Cloud Solutions) cloudy is a business and development platform that offers tools to build cloud infrastructures companies of all sizes. VK Cloud Solutions is part of the portfolio of solutions VK Digital and is based on many years of experience in the development Internet services and technologies based on. open code

2022

The emergence of the MLflow service for automating work with machine learning models

ML Platform from VK Cloud has a service MLflow for automating work with models. machine learning MLflow will help specialists to data manage the conduct of experiments, compare - ML models and their results, and reproduce tests. This was announced VK by the company on October 21, 2022.

Illustration: databricks.com

Cloudy Cloud The ML Platform combines the pre-configured JupiterHub and MLflow services to experiment with data and work with machine learning models. This is a completely ready-made environment for creating ML solutions that can be deployed in the cloud in a few clicks.

This component of the MLflow platform allows companies to implement the MLOps concept to standardize model development, reduce production time, effectively track and control versions. Using the ML Platform, data engineers and data scientists will be able to independently deploy the environment for experiments and building models, as well as scale computing power for any load. This simplifies the hypothesis testing process and increases the speed of obtaining results.

Machine learning technologies are at the heart of the most advanced products, and it is necessary that as many companies and specialists as possible can use effective ML development tools. MLflow is one of the easiest tools on the market, allowing Data Science specialists to manage the life cycle of machine learning models. At the same time, the solution works with almost any ML library, commented Dmitry Lazarenko, Product Director of VK Cloud.

On the VK Cloud platform, MLflow is integrated with JupiterHub. In the future, integration with object S3-compatible storage and Managed databases will appear.

Creating a Cloud Platform for ML Development

VK announced on August 2, 2022, the opening of access to testing the cloud platform for the full cycle of ML development. Cloud ML Platform will combine tools for experimenting with data and working with machine learning models: from setting up the environment to introducing solutions into services and business processes.

Cloud ML Platform from VK allows you to separate roles and areas of responsibility in the team, configure the task solving process and track the results of participants' experiments. Data engineers will be able to deploy the environment for working with data and building models, independently and quickly scale the computing power for any tasks and loads. Data Science specialists will be able to flexibly and quickly conduct experiments with data, formulate and test hypotheses.

Cloud ML Platform will have tools to control access to data and components of the environment, as well as to integrate the platform with enterprise user authentication, logging and auditing systems. This enables system administrators and security professionals to monitor data and quickly analyze changes to the project.

When developing a product, they relied on the use of open source components in its architecture in order to flexibly develop the functionality of the platform, to avoid dependence on the technologies of a particular vendor. Using Cloud ML Platform from VK will help Data Science teams work faster with experiments, and business - gain an advantage from the operational application of the results of hypotheses, told Dmitry Lazarenko, Product Director of VK Cloud Solutions.

At the first stage, users have access to the features of the JupyterHub service, which allows you to quickly create environments for experimenting and working with data. In the future, the functionality of Cloud ML Platform will expand, including due to the solutions familiar to data specialists based on. open code For example, platform users will be able to work with the MLflow model lifecycle management tool, container orchestration technology Kubernetes , and distributed solution. data processing Spark

VK's Cloud ML Platform will also be available to companies that prefer to work with data on their own or leased servers. The service can be deployed in the private cloud thanks to the platform's technology stack built on free software.