Profet AI

Wiwynn Drives a Bottom-Up AI Culture with Profet AI, Rapidly Cultivating AI Talent

Wiwynn, specializing in solutions for hyperscale data centers and cloud IT infrastructure, counts Meta, Microsoft, and other major U.S. cloud service providers (CSPs) among its primary customers. Benefiting from the widespread adoption of cloud applications, Wiwynn has experienced rapid revenue growth in recent years.

To provide these customers with highly efficient, low-energy consumption solutions for large-scale computing and storage, fast data transmission, and hardware-software integration, Wiwynn adopts an Original Design Manufacturer (ODM direct) model. This approach delivers customized server motherboard designs, production, and assembly services tailored to customer workloads.

“Improving efficiency and agility is very important to us,” emphasized Dr. Bey Yu, Director of Information and Digital Technology at Wiwynn. Following the company chairman’s directive from five years ago, Wiwynn has pursued a “top-down strategy and bottom-up execution” to drive digital transformation. This includes fostering an “AI-for-All” approach, encouraging employees to embrace digital tools, and establishing an End User Computing culture.

Complex Product Engineering and Resource-Intensive Quotation Process Streamlined by AI

As an ODM direct business, Wiwynn often faces labor-intensive and complex quotation processes whenever customers request quotes.

Yu explained that customer inquiries usually involve new product developments, specifying configurations and expectations while requesting quotes from Wiwynn. Upon receiving such requests, the company needs to send the specifications to all relevant units, asking them to identify and quote components meeting customer requirements. These components range from large items like product casings, metal and plastic parts, to smaller ones like cables and connectors on motherboards. The accumulated costs of these components significantly influence the final quotation. A well-done quotation results in profits; conversely, overly high quotations scare customers away, while low quotations lead to losses.

“Quotation often consumes a lot of our manpower,” Yu said. A single product quotation typically requires weeks of information gathering from multiple layers, followed by senior sales staff estimating reasonable quotes. This process is far from agile.

Thus, when Wiwynn decided to implement AI in early 2024, improving the quotation process became a key initiative. With the assistance of AI models, the relevant departments could more efficiently estimate optimal components and prices, reducing the quotation process time by 30%. Moreover, Wiwynn aims to digitize domain knowledge—previously reliant on a few senior employees—using the AI platform to accelerate talent development across departments.

Yu cited an example involving the quotation of product structural parts. One major expense in structural parts is molds, which require experienced engineers to evaluate the raw materials and processes needed to meet customer requirements. Such expertise is difficult to cultivate, with only a few “key players” in the company. Losing them could lead to significant setbacks.

Profet AI AutoML’s Ease of Use and Expert Advisory Enhance AI Analytical Skills

To significantly enhance efficiency and agility, Wiwynn introduced Profet AI’s AutoML Data Scientist Platform in early 2024. Part of the NVIDIA ecosystem, Profet AI’s platform was selected for its high usability—it requires minimal IT expertise to get started. Another reason was Profet AI’s team of consultants with extensive industry knowledge. Collaborating through AI-ML Thinking workshops with Profet AI consultants, nearly 100 domain experts at Wiwynn produced about 30 AI topics.

Although Wiwynn has trained some AI specialists, Yu admitted that these resources are insufficient to meet the demands of numerous AI projects. Furthermore, these specialists lack insight into the operations and challenges of various departments, such as manufacturing, sales, and finance. Profet AI’s consultants, with their deep industry knowledge and AI implementation experience, enable domain experts in different departments to quickly adopt AI.

“The value of consultants is in guiding teams to think,” Yu said. Profet AI’s consultants assist users in identifying areas where AI can be applied, clarifying key factors influencing AI, and pinpointing highly relevant data. Developing these analytical skills is something Wiwynn’s in-house AI talent could not achieve alone.

AI Workshops Attract Over 100 Participants, Boosting R&D Efficiency and Realizing AI-for-All Goals

Earlier this year, with Profet AI’s support, Wiwynn organized a series of AI workshops. Initially, 7–8 teams were expected to participate. However, within a month, over 20 teams and about 100 people signed up to explore AI topics for their respective departments. After months of validation, around 10 AI projects are expected to materialize.

In addition to AI applications in quotations, AI projects in engineering R&D and factories have also matured. During the workshop competition, one R&D team applied AI to thermal management. They’ve drastically reduced the time required to determine optimal parameters by half with the help of Profet AI’s platform. Another team proposed factory production line optimization, using the Profet AI platform to swiftly adjust line configurations, such as workstation layouts and machine allocations, achieving operational excellence.

Yu noted that months of AI workshops have led to increasing departmental engagement in AI initiatives. Moving forward, Wiwynn plans to recruit a second batch of AI seed talents and invest more in AI talent development. Currently, Wiwynn is collaborating with Profet AI consultants to establish a cross-departmental virtual organization called the “Center of Excellence (CoE).” This organization will facilitate knowledge sharing and experience exchange among employees involved in AI projects, supported by professional consultants to resolve challenges and accelerate AI implementation across departments.

After establishing a robust AI adoption mechanism in Taiwan, Wiwynn aims to extend this model to its global operations, fully realizing the chairman’s vision of “AI-for-All.”

As AI technology advances, its applications are transforming diverse industries. In 2024, BenQ, a subsidiary brand of Qisda Corporation, embarked on integrating AI capabilities. Recognizing the critical role of product maintenance in brand loyalty, BenQ implemented an AI strategy two years ago to optimize supply chain management. Leveraging AI to forecast future maintenance part demands enables BenQ to make accurate predictions based on historical data and market trends while dynamically adjusting models to address sudden demand changes.

“Managing repair part inventory without precise control creates significant cost pressures,” said Steven Yan, Sr. Manager of Information Technology Service Center at BenQ’s Enterprise Services Department. BenQ’s products have warranty periods ranging from two to five years, extending up to seven years for B2B clients. In the past, experienced technicians forecasted three-year repair part stock requirements based on product sales and repair data, aiming to minimize dead stock costs and prevent shortages that could increase repair costs and impact customer loyalty.

Creating a “Digital Master” for Supply Chain Management to Significantly Reduce Excess Inventory Costs

To improve inventory forecasting accuracy and ensure data-driven decision-making, BenQ opted for an AI-driven approach to supply chain management. Initially, other AI platforms were tested, but due to usability challenges and the need for extensive IT involvement, the results fell short of expectations. After exploring Profet AI’s AutoML platform, BenQ was impressed with its user-friendly interface, high model performance, and the dedicated support from Profet AI’s manufacturing-experienced consulting team, leading to a quick decision to adopt the platform.

“Profet AI’s software significantly reduces our dependence on IT staff and data scientists,” noted Yan, “empowering the supply chain department to effectively forecast future maintenance part inventory needs, achieving a forecast accuracy exceeding 80% and saving substantial inventory costs annually.”

AI Workshop Speeds Internal Talent Development

BenQ’s successful and seamless AI adoption was bolstered by Profet AI’s team-led AI-ML workshops. “Consultant support was crucial,” Yan remarked, adding that initially, employees feared AI’s learning curve would be steep. However, Profet AI’s team guided participants through practical methodology over three months, allowing employees without IT backgrounds to ideate on real business scenarios, identify key parameters for modeling, and optimize workflows. The workshops fostered a cohort of ‘citizen data scientists,’ ensuring successful AI integration within the company.

AutoML Helps Optimize Customer Profiles for More Precise Digital Marketing

After successfully resolving supply chain challenges, BenQ applied Profet AI’s advanced algorithms to its digital marketing strategy six months ago, further transforming its business.

Yan explains that BenQ previously utilized various tools to collect and analyze consumer data. These included RFM modeling to segment customers by recency, frequency, and monetary spend, Google Analytics (GA4) for tracking website engagement and conversion flows, and a Customer Data Platform (CDP) for consolidating first-party data. This data spanned customer activity across BenQ’s official website, e-commerce, customer service, and event systems, creating detailed customer profiles.

With these customer profiles now quantified into actionable insights, BenQ leverages Profet AI’s algorithms to predict product purchase probabilities. This approach enables BenQ to identify customer behavior patterns, such as paths taken by converted and unconverted customers. Once customers engage with any touchpoint on the sales platform, BenQ can anticipate purchase likelihood within specific timeframes and strategically present incentives to encourage conversion.

Following the success of AI-enabled digital marketing, Yan notes that BenQ has been expanding Profet AI’s platform to address predictive tasks in other departments. For example, Profet AI’s algorithms have streamlined product parameter adjustments in R&D, significantly reducing development time.

Looking ahead, BenQ envisions creating “digital experts” across all divisions with Profet AI’s AutoML platform, empowering each team to independently initiate AI-based modeling and testing for critical challenges. This vision aims to foster agility, drive efficiency, and support sustained operational growth.

With human life shrouded by the global climate crisis, more than 130 countries have announced that they will reach net zero emissions by 2050, and ESG has become a trending topic. Three years ago, Hsinchu Gold Bamboo² saw an uptick in interest for environmental sustainability and used bamboo particles to develop bamboo straws, bamboo tableware and electronic packaging materials among other innovative products. Last year alone, 600 million bamboo straws were shipped worldwide.

Hsinchu Gold Bamboo²’s unique technology which prevents bamboo straws and tableware from getting moldy or having other odors is the primary reason why Hsinchu Gold Bamboo² could quickly establish a foothold in the international market. Alex Shu, founder of Hsinchu Gold Bamboo², remarked that bamboo will become a natural adhesive after being grounded into powder. Over and above removing moisture at high speed, micronizing bamboo powder enables the fusion with other raw materials to become bamboo particles for application in different fields, making it an innovative raw material for environmental sustainability.

Using AI to select bamboo enables Hsinchu Gold Bamboo² to overcome restrictions on bamboo species selection

The key to successfully transforming traditional agriculture and turning bamboo into gold is to leverage AI technology. “The technology of making bamboo powder particles is not difficult, the key lies in the technology of selecting materials (bamboo),” Alex Shu gave an example, in order to reduce the carbon footprint of the products, the bamboo used in the research and development of Hsinchu Gold Bamboo² products is currently from bamboo forests all over Taiwan, However, there are 58 bamboo species among Taiwan’s 186,000 hectares of bamboo groves. These bamboos not only vary in species but also differ in moisture content due to different geographic regions, storage times, and harvest years. This introduces numerous variables into the process of developing bamboo particles.

Prior to the introduction of AI, in the selection of bamboo species for product development, the process involved manual selection. Bamboo with a strong bamboo flavor, like green bamboo, and bamboo with tough nodes, like sweet bamboo, was discarded to prevent any residual bamboo taste in bamboo straws and tableware produced. This manual selection aimed to avoid issues where the bamboo powder might not be fine enough during the bamboo grinding process, causing bamboo particles to clump and significantly reducing the selection rate of bamboo.

After introducing Profet AI’s AutoML platform, intelligent AI can modulate the best bamboo powder formula through modeling according to variables such as bamboo production area, bamboo species, moisture content and hardness, bamboo color, etc., facilitating Hsinchu Gold Bamboo² to develop bamboo powder of the same quality even if different bamboo species are used, breaking through the limitations of bamboo species and making the acquisition of raw materials for Hsinchu Gold Bamboo products more flexible.

AI modulates the best formula and process of bamboo particles, and the R&D efficiency is increased by 20% – 25%.

Besides resolving the issue of bamboo selection, Profet AI’s AutoML can also further predict the optimal formula and process of bamboo particles corresponding to bamboo powder when applied in different products, greatly shortening the R&D time and saving R&D manpower, thus improving R&D efficiency by 20% – 25%.

Alex Shu stated that bamboo powder is widely used in straws, lunch boxes, cup covers, tableware, electronic packaging materials, and medical supplies. These different products actually correspond to different production conditions such as injection molding and extrusion molding, which, in turn, require different bamboo particles. For example, bamboo particles required for electronic packaging materials can withstand higher temperatures than bamboo straws and tableware, and they also need to consider the electrostatic protection benefits and weak acid effects. In order to achieve these product characteristics, it is necessary to use bamboo powder with different formulas to produce these bamboo particles.

Profet AI successfully assisted Hsinchu Gold Bamboo² in accurately predicting material formulations. To achieve the outcomes of being produced under what conditions, the product being able to withstand certain temperatures, possess specific hardness and brittleness, raw materials have to be matched with bamboo powder in a ratio that will yield a characteristic closest to the characteristics required by the product.  In the past, when manually concocting bamboo particle raw materials for products, it was necessary to produce test strips made of 15 different raw material ratios for continuous testing in order to yield a characteristic closest to the expected product characteristics. With the assistance of AI, as long as test strips made of 5 different raw material ratios were produced for testing, the expected product characteristics could be achieved. This shortened the R&D testing cycle from two weeks to less than one week while increasing the accuracy of the raw material ratio for bamboo particles by twice that of manual concocting.

AI applications are expected to be extended to manufacturing to improve yield in the future

Introduced at the end of 2022, Hsinchu Gold Bamboo², which had been applying AI for less than a year, initially intended to use AI for customer management. For instance, they aimed to use AI models to predict warehouse inventory levels, determine when to restock specific products, and assist in setting suggested prices for new products. However, after in-depth discussions with Profet AI consultants, they realized that integrating AI into the R&D process would significantly move the needle.

Alex Shu pointed out that despite having developed many innovative bamboo products, Hsinchu Gold Bamboo² is more like an IC designer in the manufacturing industry, constructing a product development model and outsourcing it to manufacturers for production. Introducing Profet AI enabled us to avoid R&D pitfalls, quickly achieve product development goals, and improve the accuracy of product testing.

Moving forward, Hsinchu Gold Bamboo² hopes to collaborate with production manufacturers to introduce AI into the process of testing products to finished products. In addition to improving product yield, they also hope that through AI, traditional agriculture can break through existing frameworks and progress towards mass production, standardization and repeatability, and gain a new foothold in the blue ocean of environmental sustainability.

BenQ Materials, which has hit a plateau in production and manufacturing, made another breakthrough by introducing the AI automated machine learning software platform (AutoML) from Profet AI. Needs from the start of the research and development to the tail end of production management were holistically addressed, over and above enabling critical factor analysis, simulation of process parameters and rendering formula explorations easier to improve process and production bottlenecks, thereby improving the Company’s competitiveness, it also significantly reduced investment costs and brought about high efficiency.

BenQ Materials Corp. (hereinafter referred to as BenQ Materials), which actively transformed from the industrial materials industry, has successfully ventured into the medical and automotive business from the panel display industry in recent years. Medical products such as medical packaging materials, medical consumables, contact lenses, etc. have grown progressively. New items such as non-woven fabrics and functional PE films have been added to the suite of medical products this year after the acquisition of 51% equity of WEB-PRO Corp. in the fourth quarter of last year, boosting the proportion of revenue. Automotive separator film and automotive PDLC Smart Film have been introduced to the automotive market, and growth is expected in the future for the Company.

Under the strategy of diversified plan of action, besides the expansion of medical and automotive businesses, internally, BenQ Materials has also further reinforced the upgrade of technologies, actively introducing AI applications to increase product yield and optimize management and driving the overall operation performance. Led by the President, several supervisors underwent supervisor courses and set up technical courses in various departments to promote AI. The results have progressively emerged, where on top of the improvement in management performance, there was also a significant reduction in investment costs.

Cost-Effective and High Flexibility AI Tools Paired with In-Depth Domain Knowledge

Saving Eight Figures in Costs Within the Year

Fred Yu, Fred Yu, Factory Manager & AI ProjectManager, TY Factory, Polarizer Production Center, commented that it had only been a short while since the introduction of AI at the plant, and the evaluation only started around 2021, translating to only about two years time span. The key mindset then was to solve production problems. From the perspective of the production line, the aim is to raise First Pass Yield. Fred Yu pointed out that initially, two international brand systems from the United States and Japan were sourced to resolve the issues at the plant, but both systems were pulled off the plug shortly after use. The reason behind this was the discovery of a far more superior system.

Fred Yu candidly remarked, “we found a better product with higher return on investment. Profet AI products were better value for money, had more powerful functions, supported Chinese language options, can address our needs and possessed flexibility for adjustments.” Subsequently, the plant converted to the use of Profet AI and Oracle systems, and to date, more than 20 projects have been successfully implemented through Profet AI.

Fred Yu pointed out the crux of the matter, “the initial two systems were, after all, foreign brands, and it is challenging to adjust many things for one manufacturer.”

From the perspective of investment cost-effectiveness which enterprises place much emphasis on, the initial two systems from the United States and Japan would cost a total of more than NT$20 million a year. Compared with the subsequent Oracle complemented Profet AI replacement system, the latter cost approximately NT$1 million. Topped up with Oracle products, the total investment would amount up to NT$2 million. After trial calculations, Profet AI’s system this year has saved 8-figure costs for BenQ Materials, and the Company is quite satisfied with this achievement.

Fred Yu further elaborated that domain knowledge is the most important thing for manufacturing industry to build AI, but not all AI have domain expertise. For example, accuracy would be low when a large amount of data is fed into the learning model. If there is a lack of domain knowledge, there could be a possibility of caching noise, and without domain knowledge converging, cross-validation will yield differences. 

Introduction of AI successfully improved the manufacturing process, broke through the production bottleneck and carried out formula exploration

With the introduction of AI applications, what results have BenQ Materials seen so far? Fred Yu stated, “besides significant improvements in the manufacturing processes, BenQ Materials has also successfully broken through production bottlenecks.” In the past, BenQ Materials used to draw trend charts based on raw data to make informed decisions. With tens of thousands of inaccurate data on hand, it is not possible to draw these charts quickly and accurately solely by manual labor.

In addition, if the production line needs to be switched in a short period of time, the parameters need to be adjusted. The upper and lower limits of the repeated specifications of the former and latter products are very narrow. In the past, only the internal line could be switched, and this process generates a lot of waste, such as the waste of production materials. After importing the AutoML system of Profet AI, the production line adopts a non-cutover method, adjusting the concentration and temperature in advance on the other side, resolving the challenges of low manual operation accuracy, thereby reducing the scrapping of production line materials and improving yield.

Meanwhile, formula exploration was also a project which BenQ Materials benefited tremendously through AI. With so many new products to develop but so little amount of data, Profet AI’s AI tools are able to capture a similar type of data to meet the new specifications required by R&D. To boot, the open architecture of the system supports the usage status of internal and external sites, such as ESG issues such as electricity consumption, temperature, and carbon tax conversion, all of which have been well thought out.

Regular reviews of demand and function

Continuously optimize user experience to achieve win-win goals

Fred Yu said that there are two main considerations for the selection of AI systems for BenQ Materials – the system needs to meet user needs and it needs to be functional. To put it simply, things bought should be practical. The introduction of a new system is best determined by the users. If the introduction is solely led IT, and when IT and the plant production teams fail to communicate effectively, system changes will happen time and again.

One of the important reasons why BenQ Materials initially jumped ship from the original American and Japanese systems to Profet AI is that Profet AI regularly checks which functions need to be modified for customers, and the system meets visualization needs. Fred Yu emphasized that although meeting the demand and the functional needs is very basic, it is actually the most important.

Jonathan Yu, Sales Director at Profet AI, stated that during the collaboration with BenQ Materials, the R&D teams of both parties held functional R&D meetings every quarter. The Profet AI team will evaluate the actual usage needs of BenQ Materials and incorporate the needs into the future product roadmap, including topics such as data inspection, formula exploration and key factor analysis integration, device anomaly detection function optimization, and online AI model management, etc.

Jonathan Yu believes that only by continuously improving and optimizing the functions required by users, designing an operation line that is more suitable for users, and tightly integrating actual manufacturing operation processes and R&D can a win-win goal be achieved.

Who would have thought that MAXXIS, located in Changhua County, with car tires, light truck tires, motorcycle tires, bicycles among other product lines which are sold in more than 180 countries worldwide is the world’s 10th largest tire company with an annual revenue of USD 3.9 billion? At the beginning of 2023, MAXXIS, a subsidiary of Cheng Shin Rubber, became the first Taiwanese tire factory to enter the BMW supply chain as the OEM tire for BMW’s 1 series vehicles in Europe.

The bold digital transformation of MAXXIS in recent years helped win the favor of an internationally renowned car manufacturer. In 2015, in response to the requirements of multiple European car manufacturers to keep measurement records of finished and semi-finished products during product development and production manufacturing, MAXXIS rolled out their Industry 4.0 Transformation Plan and gradually automated the global production lines. The measurement data of product length, width, height, thickness, temperature, and pressure collected by point laser, line laser and 3D scanning in the machines were stored in a database.

However, as the collected data became increasingly large, MAXXIS faced the challenge of being unable to effectively analyze the massive data in the database using existing systems. After repeated evaluations, it decided to adopt the easy to use and user-friendly Profet AI AutoML data scientist platform.

Use AI to retain seasoned experts’ rule of thumb to optimize the product life cycle

Employees in the plasticization and rubber industry have always relied heavily on rules of thumb. Problems from the development of various material formulations, process improvement to raw material price procurement were solved based on the experience of seasoned experts, and MAXXIS is no exception. They introduced AutoML with an eye towards breaking through the framework by utilizing the AI tool to leverage on the experience and data of seasoned experts, build models with AI, thereby retain important experience in the Company, and continuously optimize the life cycle of products from R&D to production.

“I wanted to introduce AI, mainly because in the past, the data on the production line was mostly analyzed by people, and it was very difficult to achieve real-time feedback,” Justin Chen, Chief Information Officer of MAXXIS explained. In the past, based on different customer or market demands, the manufacturing parameters of products on the machines were adjusted by experienced technicians or senior employees. For example, it is relatively impossible to systematically analyze the data collected by the machine if the width of the tire were to be increased or decreased. After the introduction of AutoML, the models established by the AI analysis parameters can be applied to overseas factories to reduce the expatriate manpower and accelerate the expansion process over and above being able to analyze the key factors affecting the production process with the help of AI to improve efficiency.

In addition, product R&D was also a key focus of MAXXISr’s introduction of AI. “A tire is a relatively rigorous product, and the R&D cycle of a single tire requires more than two years and hundreds of raw materials,” Justin Chen remarked. Currently, the R&D department has adopted AutoML for a year and has attempted to identify dozens of factors that may affect tire fuel consumption, wet grip, and how far the tire will slide when braking, based on the hundreds of factors in tire manufacturing. Among others, type of adhesive, wire width, type, and arrangement density were key factors affecting tire performance.

Post identification, MAXXIS can use these factors to model through AI and develop tires that meet or exceed the needs of foreign car manufacturers and markets. For example, for sports performance markets, tires with high dry grip may be needed while hybrid electric vehicles require low and quiet tires.

Expand AI applications to allow departments to apprehend the data behind critical issues

After experiencing the benefits of adopting AI in production and R&D, MAXXIS began to expand its AI application in 2023, progressively introducing AutoML to more than 10 departments, such as human resources and sales, locally and abroad.

Profet AI also sent consultants to hold workshops on various AI applications at MAXXIS in an effort to assist in the rapid implementation of AI in various departments. Justin Chen stated that during the three months of training, the consultants of Profet AI started from the basic concepts of AI, before moving on to discuss with various departments, and organizing the data that could be analyzed for the difficulties and inventory problems faced by each department. The data was subsequently modeled to identify areas for improvement or optimization.

Justin Chen commented, “For example, the nurturing of key talents can greatly reduce the burden of IT personnel to adopt new systems.” With these training courses, users are able to pick up skills quickly and would also have a relatively high learning curve and willingness to learn.

Rome was not built in a day and so is digital transformation not an overnight integration. MAXXIS started from a single point, gradually expanding its AI applications from that point, line, and aspect, so that frontline employees in various departments can collect and understand the data, interpret and leverage on them. Besides accelerating the digitalization of the enterprise, it also laid a solid foundation for MAXXIS for the global market layout.