When companies work on AI development, building a development environment can be a bottleneck. Although they originally wanted to concentrate on core tasks such as model development, there are many cases where resources are taken up by environment construction such as managing computational resources and learning models, making it difficult to improve development efficiency.
Macnica 's ``AI learning environment construction service'' supports this infrastructure. In addition to system construction, the service also provides NVIDIA hardware. We strongly support corporate AI development, from technical study sessions for those involved in AI development in the field and managers who support them, to support for troubles during and after implementation.
This time, we conducted an interview with Murata Manufacturing Co., Ltd., which has introduced this AI learning environment construction service and is accelerating AI development. We asked Senior Manager Kentaro Tanabe, Manager Naoki Tokumoto, Principal Engineer Yasuyoshi Tomono, and Naoya Ito of the Data Science Promotion Division, Common Platform Technology Center, Technology and Business Development Division about the background to its introduction and usage examples.
Key points of this case study
・Increase in management costs and lack of computer resources due to increase in AI projects
・Achieve unified management of the development environment with Macnica 's AI learning environment construction service
・The deciding factor for selecting NVIDIA was its proven track record and reliability
Customers interviewed this time
From left, Mr. Yasuyoshi Banno, Mr. Kentaro Tanabe, Mr. Naoki Tokumoto, and Mr. Naoya Ito from the Data Science Promotion Division, Common Platform Technology Center, Technology and Business Development Division, Murata Manufacturing Co., Ltd.
Company website: https://www.murata.com/ja-jp
With the increase in AI projects, there is an urgent need to review the development environment
Murata Manufacturing is an electronic parts manufacturer that conducts research, development, production, and sales of ceramic-based electronic devices. The company's products are used in all kinds of electronic devices such as smartphones, televisions, and PCs, and it is a global company that boasts one of the top market shares in the world.
The company is currently focusing on promoting DX based on its long-term plan "Vision 2030," and an important element of this is AI development. However, it's not like I recently started working on DX. To begin with, the company had been introducing digital into its operations even before the word "DX" had spread throughout the world, and was proactive in data analysis and data utilization.
Within Murata Manufacturing's data science field, the Data Science Promotion Division of the Common Platform Technology Center of the Technology and Business Development Division is responsible for the development of new elemental technologies. The organization collaborates with all business departments, including the corporate department and manufacturing department, and is working on DX promotion and AI development across the company.
On the other hand, when considering future AI development, there were also issues that needed to be resolved.
"Up until now, engineers had to purchase the computer environment they needed for AI development, or use AWS or Azure. However, as the number of AI projects increases, the cost of building a computer environment increases. The company also had the problem of not being able to grasp the overall picture, making it difficult to keep an eye on the budget.'' (Mr. Banno)
In fact, the number of AI projects at the company continued to increase. If this situation continues, there is a risk that AI development will stall due to increased management costs and a lack of computer resources. Therefore, the company introduced Macnica 's AI learning environment construction service.
``Introducing large computers and sharing them can reduce costs.Also, engineers generally tend to handle development problems by themselves, but the development environment is divided between departments. Centralized management allows us to share knowledge, which speeds up resolution.With that in mind, we decided that if we were looking ahead to the future, it would be better to introduce an AI learning environment construction service.'' (Mr. Tokumoto)
Provision of NVIDIA products, high security, and generous support are decisive factors
There are three reasons why they chose Macnica 's AI learning environment construction service over competing services. First, Macnica is an official NVIDIA distributor, and the service also provides NVIDIA hardware.
"NVIDIA's data center GPU 'V100' is used for AI development in various industries including academia.If you are going to build an environment from now on, you should use NVIDIA products with a proven track record and reliability. I thought we should adopt it.'' (Mr. Tokumoto)
The second reason was the support provided by Macnica. Murata Manufacturing had previously done business with Macnica and was well aware of the extensive support the company provides. "If we're going to go to the trouble of introducing it, it's best to use a company with the best support so that things will go more smoothly," said Tokumoto, which led them to introduce Macnica 's AI learning environment construction service.
"In the early stages of implementation, there were many things we didn't understand. Macnica held a technical study session and continued to provide support, so we were able to build the environment very smoothly." (Mr. Ito) )
Finally, strong security is guaranteed. Needless to say, Murata Manufacturing is one of Japan's leading electronic parts manufacturers. Its technology is world-class, and the data it handles can be said to be of the highest level of confidentiality. To prevent such information from leaking in the unlikely event that it happens, high security is required in the AI development environment.
“Murata Manufacturing has a culture of ``Need not to know.'' Information management is very strict, so no matter how effective AI is, a large amount of data suddenly becomes available. If we were to move it to the cloud, we probably wouldn't have gained understanding within the company.'' (Mr. Tokumoto)
In this regard, Macnica 's AI learning environment creation service is provided on an on-premise basis, and ensures complete security by incorporating the latest security software.
“If this was in the cloud, I don't think it would be possible to obtain data from the materials development department or mass production factory, which handles a lot of highly confidential information.Because it was an in-house server, it would not have been possible to obtain data from the material development department or mass production factory, which handles a lot of highly confidential information. I was able to obtain it and started working on AI development.'' (Mr. Banno)
The AI developed by the AI learning environment construction service has received a great response.
The project to introduce the AI learning environment construction service began in September 2021, and full-scale use within the department began as early as the following month. The effects would later be known throughout the company, and the impetus for this was the development of AI software that analyzes images taken with an electron microscope.
"The first thing we tried in the AI learning environment we built with Macnica was AI software that can output particle size distribution and standard deviation by inputting scanning electron microscope (SEM) images. The company checked the particles and circled each particle by hand, but that method took more than 30 minutes for each particle.By using AI, We were able to automate the work, and it had a significant impact on the company." (Mr. Banno)
When we announced the results of improving work efficiency through AI at an AI forum held internally, we received a flood of inquiries from within the company. ``As the effectiveness of image analysis using AI became more widely known, there was a trend to use AI in more areas,'' says Mr. Banno.
Since the development of AI, which shocked the entire company, various initiatives have been underway at the company.
For example, in terms of images, it is an attempt to use not only 2D images but also 3D images. In order to digitize three-dimensional objects, point cloud data is used that shows position information from countless points, and this data is processed using AI.
"3D point cloud data has a huge amount of data compared to flat image data. It is difficult to handle such large data with AI in a conventional environment. However, with the AI learning environment and NVIDIA products introduced this time, It can be processed at a realistic speed.'' (Mr. Ito)
AI efforts using 3D point cloud data are still in the development stage, and full-scale utilization is yet to come. However, Mr. Ito says, ``Up until now, we couldn't even consider it, so it's great that we can now think about it realistically.''
Additionally, the development of AI chatbots using natural language processing is progressing. The system creates a database of in-house know-how and troubleshooting methods that can be called up using natural language.
``Originally, we had a rule-based FAQ for internal use.However, with the recent trend of ChatGPT, we decided that we should also work on a chatbot using natural language processing.'' (Tanabe) )
It is a well-known fact that ChatGPT's processing performance is supported by large-scale GPU servers. Even if it is limited to internal use, if you are developing a highly accurate chatbot, a rich computer environment is a must. In that sense, it can be said that the AI learning environment provided by Macnica, which allows the use of NVIDIA products, was optimal.
Murata Manufacturing’s AI strategy to realize “Vision 2030”
Currently, development using the AI learning environment is mainly carried out by some employees of the Common Infrastructure Technology Center. However, since the effectiveness of introducing AI has been proven, the company plans to continue sharing know-how and enable as many employees as possible to handle the AI environment.
We would like to aim for a state where we can take over and utilize AI in our own departments. I feel it would be nice to have a solution (NVIDIA AI Enterprise) that allows many employees to use the AI environment at the same time, provides an SDK that accelerates development, and provides support, as well as a scheduling function that allows full use of GPU resources. (Mr. Banno)
Murata Manufacturing has set out a ``three-tiered portfolio'' as a growth strategy in its ``Vision 2030.'' The first layer is ``evolution of core businesses and evolution of business models,'' the second layer is ``application-specific devices,'' and the third layer is ``creation of new business models.''
Mr. Tanabe says that the latest technology, including AI, is essential to realizing this third layer.
"The role of our organization is to use backcasting thinking to think about what is possible with technological advances and how to incorporate them into business. Technological advances are rapid, and it is difficult to predict what will happen in the future. However, it is important to have a solid foundational AI development environment in place in order to be able to get started quickly when technological advances occur.'' (Tanabe)
In the future, the utilization of AI will be essential for companies to increase their competitiveness. However, in order to develop AI in-house, the bottleneck is not only the computer environment but also the construction of the development environment. In order to quickly prepare the environment and begin model development, it is best to use Macnica 's AI learning environment construction service.
Murata Manufacturing is promoting AI development by introducing this service. The company is taking a steady step toward realizing the third layer of its portfolio in Vision 2030, which is ``creating new business models.''
Detailed information
We have prepared downloadable materials for details of this case.
Please use all means.
About the adopted AI learning environment construction service
For more information, please see the AI learning environment construction service product page.
Other NVIDIA product introduction examples
Inquiry
