THANK YOU FOR SUBSCRIBING
CIOAdvisor Apac
×
CIO Advisor APAC Weekly Brief
Be first to read the latest tech news, Industry Leader's Insights, and CIO interviews of medium and large enterprises exclusively from CIO Advisor APAC
Sanei Hytechs: Empowering Semiconductor Industry with Comprehensive Machine Learning Services
Sanei HytechsIf there is one company that truly warrants credit for the advancement of machine learning in the industry, it’s Sanei Hytechs. With over 300 employees holding extensive technical expertise, Sanei Hytechs, one of the largest semiconductor design service companies in Japan, is ready to grasp the latest trends and is constantly innovating on them. In the field of ML, Sanei Hytechs excels in providing one-stop services for diverse work processes, such as data annotation, algorithm design, and embedding in edge module devices. This is made possible through the company’s strong collaborations in China, Vietnam and Japan. Yoshihiro Mabuchi, President and CEO of Sanei Hytechs, says, “We have solid technical skills in digital design, embedded software development, and electrical circuit design. In addition, we have analog design and IoT/ICT building construction technical skill, which can prove useful for system deployment. We will also use this knowledge to provide consulting.”
As ML is gradually becoming an exploratory process, Sanei Hytechs reduces risk by helping a client understand this process and proceeding with development in certain phases.
The company has also set up an annotation center at its subsidiaries in Vietnam and China to provide Data Annotation Services for large ML datasets. Sanei Hytechs has further refined its core competency—total LSI design support—from upstream to downstream. The company is currently building an organization that can support IoT system development and the ML business.
Drawing on its extensive experience in Data Annotation services and the many achievements in Data Annotation, the company is focused on tool development. In Algorithm Design Service, Sanei Hytechs generates ML model according to the purpose—based on an algorithm as provided by a client or as published on research paper, with architecture design, transfer learning, fine tuning, pruning and distilling algorithms. The company is an expert in Imaging ML, such as Segmentation, and can improve the performance with Multimodal or Multitask methods. When it comes to Embedded Service in Edge Module Device, Sanei Hytechs optimizes embedded architecture for ML models, with quantizing, reducing parameters, compressing quantity of calculation, in GPU, Edge TPU, VPU, MPU, FPGA, and more.
The value that Sanei Hytechs brings to its clients can be exemplified with the company’s engagement with a client that provided an algorithm conception. The team of experts at Sanei Hytechs programmed the algorithm in Keras and evaluated the inference accuracy. They restructured the architecture to improve accuracy and cleared the accuracy target. At the same time, the processing speed and logic circuit scale were estimated, and improvements were examined. “The circuit scale has been reduced by pipelined, resource sharing, and memory access reduction, and the amount of calculation has been reduced by formula conversion to improve processing speed,” notes Yoshihiro Mabuchi.
Since 2015, Data Annotation Service has resulted in a wide variety of achievements for automobile-related companies in Japan. Sanei Hytechs’ strength lies in being particular about inspecting all products and leveraging its team’s know-how to maintain high quality. Furthermore, by linking with algorithm design, it is possible for the company to deliver accurate advice from an algorithmic point of view. Yoshihiro Mabuchi adds, “In algorithm design, we not only pursue performance, but also derive the optimum solution in cooperation with Embed in Edge Module Devices that can target various edge devices. We will keep on trying to gain great success without the fear of change, and with a superior sense of balance, we will provide a high degree of specialization.”
Drawing on its extensive experience in Data Annotation services and the many achievements in Data Annotation, the company is focused on tool development. In Algorithm Design Service, Sanei Hytechs generates ML model according to the purpose—based on an algorithm as provided by a client or as published on research paper, with architecture design, transfer learning, fine tuning, pruning and distilling algorithms. The company is an expert in Imaging ML, such as Segmentation, and can improve the performance with Multimodal or Multitask methods. When it comes to Embedded Service in Edge Module Device, Sanei Hytechs optimizes embedded architecture for ML models, with quantizing, reducing parameters, compressing quantity of calculation, in GPU, Edge TPU, VPU, MPU, FPGA, and more.
The value that Sanei Hytechs brings to its clients can be exemplified with the company’s engagement with a client that provided an algorithm conception. The team of experts at Sanei Hytechs programmed the algorithm in Keras and evaluated the inference accuracy. They restructured the architecture to improve accuracy and cleared the accuracy target. At the same time, the processing speed and logic circuit scale were estimated, and improvements were examined. “The circuit scale has been reduced by pipelined, resource sharing, and memory access reduction, and the amount of calculation has been reduced by formula conversion to improve processing speed,” notes Yoshihiro Mabuchi.
Since 2015, Data Annotation Service has resulted in a wide variety of achievements for automobile-related companies in Japan. Sanei Hytechs’ strength lies in being particular about inspecting all products and leveraging its team’s know-how to maintain high quality. Furthermore, by linking with algorithm design, it is possible for the company to deliver accurate advice from an algorithmic point of view. Yoshihiro Mabuchi adds, “In algorithm design, we not only pursue performance, but also derive the optimum solution in cooperation with Embed in Edge Module Devices that can target various edge devices. We will keep on trying to gain great success without the fear of change, and with a superior sense of balance, we will provide a high degree of specialization.”
