Memorence AI: Digital transformation with learning capabilities

How can small and medium enterprises that lack data apply artificial intelligence (AI) to digital transformation and automate their operations? Memorence AI, a startup made it to the list of top 10 teams joining the final round of 2021 Qualcomm Taiwan Innovation Challenge and a silver medal winner of 2021 APICTA awards under the category of "industrial application – manufacturing and logistics," discovered an answer to that question.

Memorence AI founder and CEO Pai-heng Hsiao told DIGITIMES that the company's goal is to create an intelligent learning system that emulates the human brain, turning human knowledge and operational experience into digital knowledge, and applying it to digitize and automate manual operations so that AI can become a daily productivity tool that everyone can easily use.

"We break down the process into several modules: data capture, knowledge generation (i.e. image or data tagging), AI learning and continuous updating, and the whole process is strung together," Hsiao explained that is how they help large enterprises and SMEs to become digital and intelligent.

For example, a food processing plant uses traditional machines to cut fruits and vegetables, but the machines are not very accurate and may waste raw materials. Memorence AI developed a proprietory algorithm to improve cutting performance. "We define ourselves as a cross-domain core AI provider, even for different industries," said Hsiao, who stressed that through their cross-domain AI learning system, a client firm can generate digital assets, digital knowledge and AI models, allowing clients to digitize their processes, improve quality and efficiency, reduce labor costs, and improve the quality of manual operations.

Hsiao did his thesis on the topic of "From the neocortex structure of the brain system to the compiler and to the design of neural network chips for image recognition", and has accumulated over 10 years of experience designing image chips in a sensor company, as well as doing AI research work in firms and Academia Sinica, which is Taiwan's top national research institution.

Memorence AI's team size is 5-10. Hsiao is in charge of algorithm optimization, AI core algorithm, and architecture design. His teammates take care of business development, marketing, and software development. One of them has more than 20 years of experience in sensor-image processing, and he is mainly responsible for image algorithm development.

Differentiating from general AI competitors

General AI start-ups can quickly conduct proof-of-concept projects, but the prerequisite is a large amount of data. However, most companies' pain points are under-digitization, a lack of equipment and systems, and a lack of data. Hsiao said Memorence AI addresses those pain points by helping manufacturers to integrate AI into manual operations and processes, and by integrating data acquisition into the systems.

"In addition, there are some electronic companies that have had data acquisition systems, but their algorithms are not very advanced, so they have recently started to adopt AI," Hsiao said more than 90% of enterprises need this digitalization process, but the key is the integration of the systems.

Even for similar applications, different enterprises need different identification systems due to different application specifications, so it is difficult to use the same system to apply to all those problems. "How to keep the system evolving, updating, learning in real time, and linking to the original system within the enterprise, such as ERP/MES, are all important factors to make the enterprise customers buy the system," said Hsiao. "Our solution is to provide APIs for them to do the linkage."

Memorence AI became a partner with SAP in 2021 and will work together to expand markets in Singapore, China, and Taiwan. "They have very mature MES and ERP systems, and our products are directly linked to their systems through APIs, which can increase their data acquisition and AI applications in different fields," Hsiao said Memorence AI also participated in Qualcomm's QITC 2021 in the hope that it might establish a similar partnership with Qualcomm in the future.

Hsiao explained that Memorence AI's previous edge devices did not use mobile wireless (4G/5G), but then they realized that some applications required human operators to execute operations while outdoors - for example, a cell phone application must be usable outdoors. "To reduce manpower, we are considering the use of 5G mobile communication, and Qualcomm has just the right product for this. That's why we signed up for Qualcomm's New Innovations Competition," said Hsiao.

Memorence AI uses Qualcomm's latest RB5, a computationally powerful edge device that integrates a variety of sensor interfaces and AI with 5G modules. Hsiao said some applications take place indoors, such as in factories, while others are outdoors, and Qualcomm products are suitable for both, complementing the need for mobile wireless communications. At QITC, Qualcomm provided development tools and modules to help start-ups validate their creative ideas and solutions on Qualcomm's platform. They also help match up QITC teams for potential partnerships.

In terms of business models, if a hardware device is involved, such as an imaging system or a computing unit, Memorence AI will sell it to the customer and charge a one-time installation fee for the AI software and a monthly subscription to the service of the AI platform.

"This year, we have a fundraising plan for long-term product and technology development, but we hope to choose a strategic partner who can help expand our markets and share our long-term values and goals," said Hsiao.

Source: Memorence AI, 2022/3

Memorence AI founder and CEO Pai-heng Hsiao
Photo: DIGITIMES