Victor Hoon Lee, CEO of A.I.MATICS, said in a recent interview with Unicorn Factory, MoneyToday’s startup media platform, that vertical AI will ultimately create greater value in industrial environments by understanding the specific context of each industry.

“If general-purpose AI is a model that knows a little about everything, vertical AI is a specialist that deeply understands a particular industry,” Lee said. “In real-world industrial settings, AI that understands the context of the field will ultimately create more value than general-purpose AI.”

A.I.MATICS began as an in-house venture at Hyundai Motor Company in 2000. The company initially grew by developing AI-powered driving safety solutions that use in-vehicle and external camera footage to detect more than 30 types of risky driving behavior in real time, including drowsy driving, mobile phone use, driver inattention, and traffic signal violations.

Based on driving data, the platform also generates individual driver safety scores and provides personalized coaching and fleet monitoring services, helping corporate customers prevent accidents while reducing insurance premiums and operating costs.

From Hardware Subcontracting to an AI-Driven Business

Moving beyond its traditional automotive-focused business, A.I.MATICS has developed its own vision AI models and expanded into manufacturing, mobility, healthcare, and security.

“In the past, we developed technologies such as the world’s first lane departure warning and color recognition systems, but there were limits to how much we could grow within a subcontracting structure dominated by large corporations,” Lee said. “We shifted our business toward building independent competitiveness by integrating our existing technologies with proprietary AI models.”

Today, A.I.MATICS provides vision AI inspection solutions for manufacturing, facial recognition-based access control systems for security, and enterprise AI chatbots for customer support. The company is also developing an AI solution for healthcare that analyzes electrocardiogram (ECG) data to detect arrhythmias.

At the core of these solutions is aimNet, A.I.MATICS’ proprietary AI model.

One of aimNet’s key strengths is its ability to run efficiently on cost-effective edge processors. Instead of relying on NVIDIA GPUs that can cost more than KRW 20 million (approximately US$15,000), the model is optimized to operate smoothly on low-cost chipsets from companies such as Ambarella and Rockchip, some priced at just tens of thousands of Korean won.

A.I.MATICS points to its extensive real-world data as another major source of competitive advantage. The company became the first in Korea to receive an A-Class data quality certification for unstructured data from an organization affiliated with the Ministry of Science and ICT.

“We have collected more than 11 million driving and accident video clips from approximately 10,000 devices,” Lee said. “We break the footage down into one-second and half-second intervals to analyze the context before and after each event. Our ability to continuously accumulate field data and use it to improve AI accuracy is one of our most important technological assets.”

Targeting the “Last 10%” of Manufacturing Automation

One of the areas A.I.MATICS is now focusing on most heavily is the smart factory market.

“China has already implemented so-called ‘dark factories,’ where production lines can operate with the lights off and only one person overseeing as many as 10 lines,” Lee said. “We expect the trend toward unmanned manufacturing to spread to first- and second-tier suppliers in Korea within the next three to four years.”

A.I.MATICS is using production lines operated by its largest shareholder, Dreamtech, and affiliate Namuga as real-world testbeds, repeatedly validating its AI models and accumulating manufacturing data.

The company is particularly targeting what it calls the “last 10%” of manufacturing automation.

While much of the production process has already been automated, final visual inspection still often depends on human workers to identify scratches, missing components, assembly defects, and other quality issues.

A.I.MATICS says its AI vision technology can detect microscopic defects as small as 3 micrometers (μm), while achieving a 0% missed-defect rate and an over-detection rate of 3% or less.

The company is also developing an integrated system that connects AI-powered visual inspection with autonomous mobile robots (AMRs), enabling inspected products to be automatically transported to the next stage of production.

“Production lines at major corporations are already highly automated, so there are relatively few areas left for AI to newly penetrate,” Lee said. “First- and second-tier suppliers that lack in-house AI development capabilities actually represent a much larger market and a more significant pool of potential customers.”

“The final frontier in manufacturing cost reduction is ultimately labor,” he added. “Our goal is to automate everything from component transportation to final visual inspection, reducing labor costs while improving productivity.”

Source: MoneyToday (머니투데이)