A.I.MATICS CEO Hoon Lee recently sat down with Seoul Economic Daily to discuss the company’s evolution from AI-powered mobility safety into industrial Vision AI. In the interview, Lee shared how A.I.MATICS is building on its proprietary Vision AI model, aimNet, to expand from AI driving safety solutions into manufacturing with AIM-T1, while outlining the company’s broader ambition to become a global Physical AI company.

Samsung Electronics reduced insurance premiums for its employee commuter buses to just 32% of their previous level after installing AI-powered cameras on approximately 2,200 buses nationwide in August 2024. The key was artificial intelligence embedded in the cameras. Front, rear, and side-mounted cameras continuously analyze road conditions in real time, detecting potential hazards and alerting drivers before they escalate.

The cameras do more than monitor the road. An in-cabin camera observes the driver, detecting signs of drowsiness or distraction. When the system identifies risky behavior, it issues audible alerts and voice warnings. By making drivers aware of potential risks both inside and outside the vehicle in real time, the system helped foster safer driving habits—and ultimately contributed to lower insurance costs.

At the center of this system is Roadscope 9, a flagship product from Vision AI company A.I.MATICS. The device is powered by aimNet, the company’s proprietary Vision AI model.

“Approximately 6,300 vehicles equipped with A.I.MATICS’ AI-powered driving safety management technology are currently operating across Korea, including Samsung Electronics commuter buses, Air Liquide Korea tanker trucks, airport buses, and intercity buses,” CEO Hoon Lee said in an interview with Seoul Economic Daily on July 22.

A.I.MATICS began in 2000 as Hyundai Motor Group’s second in-house venture and was spun off as an independent company in 2003. Following an equity investment by smartphone component developer Dreamtech in 2023, A.I.MATICS became an affiliate of Dreamtech.

Since its early days as an in-house venture, the company had developed road-image recognition devices and supplied them to Hyundai Motor and Kia. In 2019, the company changed its name from PLK to A.I.MATICS and began focusing more extensively on Vision AI. Lee has led the company since 2020, when it began exploring new growth opportunities.

His appointment as CEO coincided with a critical turning point. Hyundai Motor and Kia began sourcing road-image recognition components from overseas suppliers rather than A.I.MATICS. The company responded by moving away from a business model dependent on a single automotive component and instead focused on developing more scalable technologies.

The result was aimNet, A.I.MATICS’ proprietary Vision AI model.

Holding up three fingers, Lee summarized aimNet in three words: “small, powerful, and cost-effective.”

Because aimNet was designed to run on compact devices installed in vehicles, the model itself had to remain lightweight. Although it is optimized for a focused set of 35 tasks, that specialization enables significantly higher computational performance for those specific applications.

According to A.I.MATICS, the latest version of aimNet delivers approximately 50% higher recognition accuracy and operates up to 22 times faster than YOLO, a widely used general-purpose computer vision model.

Its lightweight architecture also enables aimNet to operate on relatively inexpensive AI chips.

“General-purpose Vision AI models typically require high-performance computing resources, but aimNet can deliver sufficient performance even on an Ambarella NPU costing around KRW 20,000,” Lee explained.

In 2023, A.I.MATICS introduced its Roadscope series of AI driving safety devices and the aid analytics platform, both powered by aimNet.

Roadscope uses vehicle-mounted cameras to continuously monitor the driver and surrounding road environment, immediately issuing alerts when it detects potentially dangerous conditions. The aid platform analyzes driving records and provides a driver-scoring service that evaluates driving behavior and delivers feedback designed to help prevent accidents.

“In 2024, we conducted a six-month pilot project with the Korea Transportation Safety Authority involving 500 route buses, and the results showed a 71.2% reduction in traffic accidents,” Lee said. “Building on these proven results, the business generated more than KRW 4 billion in revenue last year alone.”

He added that A.I.MATICS plans to deploy the solution to urban and intercity buses in Incheon, North Jeolla Province, Jeju, and Seosan during the second half of this year.

Having established a foothold in AI-powered driving safety, A.I.MATICS found the next application for aimNet on the factory floor.

The result is AIM-T1, an automated visual inspection system developed in 2024. Cameras capture images of product surfaces, while Vision AI identifies defects such as scratches, plating defects, and assembly errors. The system goes beyond defect detection by automatically separating defective products within the equipment.

AIM-T1 achieved its first successful production deployment in December 2024, when it was introduced on selected production lines at Dreamtech’s manufacturing facility in Vietnam.

“Just as our cameras can detect driver drowsiness, AIM-T1 detects defective products. It is a Vision AI system designed specifically for manufacturing,” Lee said. “Many manufacturers still rely on human eyes and hands for final visual quality inspection. AIM-T1 provides a way to automate that process.”

According to the company, deployment on Dreamtech’s production line resulted in a near-100% defect detection rate while increasing productivity on the line by 26%.

A.I.MATICS now plans to use the Dreamtech deployment as a reference case to expand AIM-T1 sales to overseas manufacturing operations. Last month, the company established its first overseas subsidiary in Bac Ninh Province, Vietnam.

“We established the Vietnam subsidiary to strengthen sales to global manufacturers based on the results we have already validated on Dreamtech’s production lines,” Lee said. “Once our Vietnam business is firmly established, we plan to target the North American market in the second half of next year.”

Lee sees the visual inspection business as a key turning point in A.I.MATICS’ transformation from a mobility-focused company into a Physical AI company.

“Revenue from our visual inspection business was around KRW 2 billion last year, and we expect it to more than triple this year,” he said.

“At the same time that we expand AIM-T1, we are also developing our strategy for autonomous mobile robots, or AMRs. We plan to develop AI models for AMRs based on the Vision AI technologies we have accumulated over the years.”

The longer-term goal is to connect AIM-T1’s inspection results with automated equipment and AMRs, enabling automation across the manufacturing process—from identifying and sorting defective products to transporting materials within the factory.

Asked about the company’s long-term ambitions, Lee said, “We want to firmly establish our identity as a Physical AI company and have our technological strengths recognized in global markets.”

Lee identifies three core technology pillars for Physical AI: vision, memory, and control. He believes A.I.MATICS has already established globally competitive capabilities in the first two.

“We already have the ability to perceive the physical environment through cameras and to build memory from the data we collect,” Lee said. “Our goal is to take the AI technologies we have refined on roads and factory floors and expand them into areas such as healthcare, ultimately building A.I.MATICS into a Physical AI company that operates across a much broader range of industries.”

Source: Seoul Economic Daily