Hiroshi Fujimura, Ph.D., R&D General Manager

Born in Aichi Prefecture, Japan, Hiroshi received his M.S. in Media Science from the Graduate School of Information Science, Nagoya University in 2005. The same year, he joined a major general electronics manufacturer’s laboratory as a researcher. Obtained his Ph.D. degree from Hosei University Graduate School of Information Science and Technology through the Early Graduation Course in 2022, and in August, he joined AWL, Inc. as a R&D general manager, leading AWL’s technological development.

*Interview, article, and photography by the external team.

Artificial Intelligence in Business: Sympathize with business where AI technology felt close at hand.

After completing graduate school, Hiroshi Fujimura worked for one of Japan’s leading general electronics manufacturers, where he focused on speech recognition, a field covered by many AI technologies.

“By experiencing the planning and management role in the lab, I began to think that I want to work in a position where I could have a bird’s eye view of the company. Apart from that, my wife started her own business, which encouraged me to switch jobs. I was motivated to follow in her footsteps and pursue what I wanted to do.”

It was under these circumstances that Hiroshi came across AWL, Inc.

“I was looking for a company that had AI technology a mainstay of its business, where I could feel a contribution to problem solving in my immediate surroundings. AWL was the company that perfectly met these requirements.”

No matter how highly functional an AI system is, it is meaningless if it is not put to practical use and used by the public. We develop and implement camera solutions using Edge AI (technology that incorporates AI on the hardware device itself) in a variety of forms to suit our customers’ needs.

During the interview, I was attracted to the phrase “AWL is a technology company,” which I heard from CTO Tsuchida. I sympathized with the way business is done at AWL, where the results of work can be felt close at hand.

Accumulated Proprietary Know-how to Make AI even Smarter Automatically and Continuously.

AWL’s representative services are “AWLBOX” and “AWL Lite.” AWLBOX makes it possible for existing security cameras installed in retail stores to image-based intelligent solutions and offers various features tailored to your needs, such as customer analysis, visualization of flow lines and notification of customers in need of service. Another edge AI application, AWL Lite provides accurate real-time attention analysis and audience measurement to maximize the value of advertisement at an affordable price by simply installing to any digital signage.

We have also developed “AWL Trainer” to run MLOps (Machine Learning Operations) for these Edge AI-equipped services, especially for continuous learning and automatic deployment. That is how Edge AI automatically adapts itself to different situations.

“In the world of edge AI, most cases use a single general model that can be adapted to various situations and continue to use the system once it is released. There is a very high hurdle to overcome the barrier of practical application and move into the operational phase.”

An example is that in the past, “when it comes to masks, white was the standard,” but nowadays there are many colorful products, so unless the AI is updated, the recognition accuracy of face detection will become less precise. They also encounter many situations where the AI does not recognize as much as they expected when they introduce and run it.

AWL’s Edge AI functionality is used in a wide variety of locations, including cash registers, product shelves, and the front of refrigerators. The people and backgrounds captured by the cameras vary from place to place. It is extremely difficult to maintain a high level of accuracy while making AI smarter automatically and continuously for changes in its scene and time.
While many companies around the world have failed to break through this barrier, we have been persistently repeating trial-and-error and building on our proprietary know-how to construct the AWL Trainer, making practical use of edge AI. Its low cost and ease of implementation also make it easy to be accepted by many customers.

In my original field of expertise, speech recognition, there are technologies such as internet-connected cloud-based systems that change vocabulary depending on the situation and usage. Yet, I was amazed that such an Edge AI-based system could be realized.

Corporate Culture that Successfully Blends a Diverse Group of Young Foreign Engineers with Senior Talents

Hiroshi is committed to creating the next generation of AWL technology. One of his tasks is to further improve the accuracy of “AWL Trainer”. He says that the fundamental idea of the speech recognition field which he originally specialized in, is similar to that of AWL Trainer, as the model is adapted and changed for each situation. On the other hand, in addition to developing services that already have a base such as AWL Trainer, one of their missions is to create something completely new that will be the next centerpiece.

“My role is to select and execute the next R&D targets, manage resources, and cover other areas where we are lacking in R&D, while keeping an eye on customer needs and global trends. AWL has many young foreign engineers who are highly motivated to work. Meanwhile, seniors like us are making good use of our experience in R&D as a corporate researcher and communication with customers.”

When one thinks of start-ups, we tend to imagine the momentum and energy of young members, but AWL is a fusion of both young and senior talents to keep the business running strong. While in the expansion phase, the senior staff’s ability to appropriately manage the business and personnel seems to be shining through.

“We are very close to our customers, even our R&D department has meetings with them often. We have great opportunities to intervene in decision-making, and I enjoy the feeling of leading a project to success with all members of the team.”

Hiroshi’s goal is a world in which Edge AIs can become smarter by interacting and collaborating with one another while protecting the privacy of the customer. For example, an AI can automatically determine, just after installation, that “since the environment A in which this AI operates is similar to another environment B, let’s run it by imitating the one running in environment B,” and can be used immediately at the new site. The functionality is only available at AWL, where Edge AI has been adopted at numerous sites.

“In the future, when the working population decreases, I would like to use cutting edge AI technology, which can be easily integrated into familiar places, to develop products that can help and enrich people’s lives. I believe that AWL, with its advanced technology in this field, can make this dream a reality,” he said.