December 4, 2020, Tokyo – NTT DATA Corporation, a leading digital business and IT services provider, has launched a Machine Learning Operations (MLOps) introduction service that shortens the AI (artificial intelligence) service creation cycle for customers, and provides continuous improvements. The introduction service is being offered beginning December 4, 2020.

MLOps is the Machine Learning (ML) version of DevOps¹, a general concept in which the ML developer and system manager collaborate for a smooth process, from implementation to commercial system operation. NTT DATA currently provides a DevOps service aimed at quickly providing value to customers, and in response to the increasing volume of inquiries regarding new business creation using AI, has now expanded the scope of this service to include AI. Because of the need for a high degree of accuracy in AI, and continuous improvement in that accuracy following the service launch, NTT DATA is drawing on its basic know-how from the existing DevOps service for quickly delivering value to customers, and applying the knowledge gained from numerous AI projects. By offering this service, NTT DATA will be able to shorten the AI service creation cycle from several months to a few weeks, and support continuous improvement in business value.

Going forward, NTT DATA will expand the MLOps introduction service to all its group companies, aiming for 100 implementation projects within three years.

Background

Projects that attempt to use AI to establish new services or improve existing operations are increasing year by year, but in many cases these never reach full-scale implementation for actual business. There are two main reasons. First, during the limited proof of concept (PoC) stage, the accuracy of the AI fails to reach a level that supports actual use. Second, even when the AI model created by the ML developer is introduced into the commercial system by the system developer, in many cases the introduction takes excess time due to communication and division of cost issues.
Further, since the behavior patterns of people are now changing on a daily or weekly basis because of the coronavirus crisis, in many cases the AI models tuned by data scientists prior to the change experience deterioration in accuracy, and are no longer useful.
Considering these circumstances, NTT DATA decided to make a major update to its MLOps service AICYCLE®, part of the Digital Success Program² launched in March 2020, and began providing this introduction service on September 30, 2020.

Overview (Features)

The MLOps introduction service was developed based on NTT DATA's know-how from the existing DevOps service for quickly delivering value to customers, and applying the knowledge gained from numerous AI projects. Utilizing the introduction service will allow for greater efficiency in AI development, and shorten the time required for commercialization of AI services.

• Tool chain for greater efficiency in the AI development process
  With AICYCLE, the prediction model that serves as the decision logic when the AI makes a prediction is evaluated and updated automatically using various business-related data, as well as data from the results and performance of AI predictions (quality of predictions and results). Up to now, this service has used technology to maintain prediction accuracy (the quality of the prediction model). This latest version adds verified tools based on open source software (OSS) and public cloud services to Altemista^®3. Using this tool accelerates the six work processes listed below by more than 250%.

Fig. Tool chain provided by MLOps service

Business Utility of the MLOps Implementation Service

1) Generation of actual business following PoC success
In applying AI to actual business operations, during the proof of concept (PoC) stage the predictive accuracy needs to reach a level applicable for business. The key to achieving this is repeating the six processes in the above chart, and determining the optimal tuning. By accelerating this process more than 250%, NTT DATA is able to complete, with a standard three-month PoC period, the five rounds of tuning typically considered necessary for a successful trial. This leads to generation of actual business success.

2) Continuous improvement during actual business operations
The value of a business utilizing AI depends mainly on the prediction accuracy, and this accuracy can be affected by changes in information or conditions over time. For example, as a result of the coronavirus crisis during the current fiscal year, retail and restaurant sales declined sharply compared to predicted values. Even if predictive models are tuned to match such circumstances, the same issue will occur once the emergency declaration is lifted or public opinion shifts, and sales increase sharply relative to predicted values.
During such periods of rapid changes in conditions, utilizing the tool chain outlined above allows the model to be improved simply and quickly for appropriate predictions, allowing for business improvements.

Future Goals

NTT DATA will expand this MLOps introduction service to all its group companies, aiming for 100 implementation projects within three years.

Notes
1.DevOps is the establishment of a structure for close cooperation between software development (Dev) and IT operations (Ops) for prompt implementation and updating of software.

2.The Digital Success Program is a practical business transformation program, based on experience with business transformation through AI and data utilization at more than 500 global companies, and support for AI and data democratization.
See the news release "NTT DATA investing in DataRobot to accelerate their DX Services" (March 18, 2020) https://www.nttdata.com/global/en/media/press-release/2020/march/ntt-data-investing-in-datarobot-to-accelerate-their-dx-services
See the news release "NTT DATA investing in Snowflake to Accelerate DX Services" (October 7, 2020) https://www.nttdata.com/global/en/media/press-release/2020/october/ntt-data-investing-in-snowflake-to-accelerate-dx-services
 
3.Altemista is the brand for NTT DATA's suite of solutions for speedy service planning development, supporting startups and innovation emergence.
https://abler.nttdata.com/solution/altemista.html (Japanese)

*"Altemista" is a registered trademark of NTT DATA Corporation in Japan.
*"AICYCLE" is a registered trademark of NTT DATA Corporation in Japan.
*"Digital Success" is a registered trademark of NTT DATA Corporation in Japan.
*"Amazon Web Services", "AWS" and the Amazon Web Services logo are trademarks of Amazon.com, Inc. or its affiliated companies in Japan and other countries.
*"Microsoft Azure" and "Azure" are trademarks or registered trademarks of Microsoft Corporation in in the United States and other countries.
*"Google Cloud" is a trademark of Google LLC.
*Other names of products, companies, and organizations are trademarks or registered trademarks of those companies.

Contacts

Media inquiries
NTT DATA Corporation
Public Relations Department
Contact : pr-support@kits.nttdata.co.jp

Product and service inquiries

NTT DATA Corporation
Technology and Innovation General Headquarters
System Engineering Headquarters
Agile Professional Manager
Contact persons: Miyadate / Yamaguchi / Sasaki
E-mail: apcj@kits.nttdata.co.jp

Business Solutions General Headquarters
AI & IoT Department
Contact persons: Kato / Kobayashi / Miyaji
E-mail：aiiotmarketing@kits.nttdata.co.jp

News Releases.

The services, prices of products and services, specifications, telephone numbers, etc. for inquiries and other information included in news releases are the data available on the day of the release. This information may be changed at any time without notice. In certain circumstances, due to various risks or unexpected occurrences, actual results may also be different from the plans or projections in news releases.