Cloud MTS

MLOps Platform features

Convenience and flexibility of deployment, development, training and model testing

Quick launch of ML/AI solutions development without extra environment preparation and configuration costs

Customizable and reproducible workflows

Development resources control and management

Get advice

Tasks to solve

Automation

MLOps pipelines development to automate workflows throughout the MLOps lifecycle

Collaboration

Collaboration of engineers and other teams in the secure environment

Reproducibility

Tracking data and models versions, saving all project assets and artifacts

Future add-ons

Integration of multiple frameworks and tools in the course of development and growth

Scaling

Availability of a wide range of tools: from small to large

Technical features

Management of MLOps Platform atop of Kubernetes

Work in the cloud, on-premises, mixed infrastructure

Commercial and open source tool support

Tools and integrations

Frameworks	TensorFlow, Keras, PyTorch, Catalyst
Experiment Tracking	TensorBoard
HyperParameter Tuning	NNI
Remote debugging	VSCode
Distributed learning	Horovod, PyTorch 1.5, Internal Native
Model Deployment	Seldon, Cortex, TorchServe
Training Pipelines	Apache Airflow, Internal Native
Data Registry	DVC
Model Registry	Internal Native
IDEs	Built-in: Jupyter® Lab / Jupyter® Notebook; Use on-premises: VSCode, PyCharm, other familiar IDE

Zoom Get advice

Choose a convenient way to work with the platform

Command interface (CLI)

Web interface

Python API

Beta testing

Service Manager

Stanislav Obidin

The head of the MTS ML\AI service direction.
Develops the GPU SuperCloud direction, MLOps platforms and services based on artificial intelligence and machine learning.
Over 20 years in IT, including the software development field.

Advice on your project