Kim Deuk-hwa, CEO of Funzin — The Next-Generation Battlefield Led by Artificial Intelligence

Flavius Vegetius Renatus, an ancient Roman barrister, once said, “If you want peace, prepare for war.”
In today’s world, where being unarmed invites invasion, warfare is no longer decided by swords but by artificial intelligence (AI) technology.

As CEO Kim Deuk-hwa said, “Humans no longer need to enter battle,” the era of AI-driven robotic warfare—once only seen in science fiction—has now arrived. Israel has deployed the self-destructive drone Harop, which autonomously identifies and attacks its targets, while Ukraine uses machine learning-based quadcopters capable of striking enemies without human intervention. Before the world could even come to terms with these self-destructive “killer robots,” Korea began developing its own large-scale robotic combat unit under the Amitiger project to avoid falling behind. Funzin is developing the AI that will command this robotic force.

Funzin’s flagship solution, Kill Web Matching (KWM)—developed in collaboration with the Korean Army—is an AI-based Command Decision Support System, often called an “AI staff officer.” The AI identifies and evaluates targets, then recommends optimal combinations of weapons, mission plans, and attack strategies based on various battlefield conditions. Once approved by a human commander, the operation is executed immediately by robotic weapons. The entire process—from target recognition to mission execution—takes just two minutes.

“We don’t launch a nuclear strike to destroy a single enemy tank,” Kim said. “Our AI learns the commander’s mindset and recommends the most efficient tactics.”

He added, “Typically, six officers are deployed at one command post, and our AI now acts as the seventh officer. In the past, by the time the staff finished their discussions, the battle would already be over. It’s frightening to think that the enemy could also be using kill web matching technology.”
His expression, as he explained the technology, was filled with quiet confidence.

In June last year, Kill Web Matching successfully completed a test operation as part of the Boost Project for Amitiger’s development. It is now undergoing combat tests at KCTC. Once commercialized, the system is expected to command robotic weapons such as drones and unmanned combat aircraft.

However, for Kill Web Matching to function effectively, both the target recognition AI and the tactical recommendation AI must achieve high precision. Data is the core of AI, but since Korea has not experienced a real war since the Korean War, there is very little enemy-related data available. Moreover, due to security restrictions, private companies have limited access to military data. Funzin had to find a way to overcome this limitation.

To improve the performance of its target recognition AI, Funzin adopted Few-Shot Learning technology, enabling AI training with limited datasets, and further refined it for military use. The company also developed its proprietary EagleEye platform, which uses Generative AI to produce military-grade synthetic data. The product is currently being supplied to the Army.

“Even if we only have 100 or 1,000 real images, we can generate 10,000 by synthesizing and augmenting them with generative AI,” Kim explained. “For example, a daytime photo can be converted into a nighttime one, or a sunny image into a rainy-day scene. Target recognition AI trained with EagleEye’s synthetic data performs about 90% better than with real data alone. It dramatically reduces the time and cost of data building while maintaining near-perfect accuracy.”

Funzin is also advancing toward Physical AI, a logic-based form of artificial intelligence that understands the physical laws of the real world. Unlike generative AI that focuses on imagination, Physical AI produces more practical and reliable results suitable for real environments. This AI can quantify battlefield conditions more accurately and make more rational operational decisions by evaluating which tactics best fit the situation.

Still, without real battlefield data, AI cannot learn as if it were in a live combat environment. When asked about this limitation, Kim said, “That’s why we provide not only AI solutions but also MLOps (Machine Learning Operations) systems that help users operate them effectively in diverse environments. The military can train our AI models with its own data and quickly deploy or update them in the field.”
This, he said, is what makes Funzin a true system integration company.

Mobile Communication Expertise That Shines on the Battlefield

Funzin was not originally a defense company. Founded in 2006 as a mobile communications firm, it initially focused on intelligent IoT (AIoT) technologies and entered the defense market less than three years ago. Yet, the company completed a prototype of Kill Web Matching in just six months. Its rapid development drew industry attention, and by last year, Funzin had secured seven defense contracts. Recognizing its potential, Korea Aerospace Industries (KAI) invested ₩13.3 billion (10.27%) to become the second-largest shareholder with a 20.27% stake. Today, over half of Funzin’s revenue comes from the defense sector.

Kim attributes this growth to the company’s long-standing expertise in communications technology.

“Funzin has deep roots in telecommunications and IoT,” he explained. “We have strong experience with on-device environments and embedded systems that allow software to run directly on local devices instead of the cloud.”

He continued, “AI designed only for stable, resource-rich environments—like the web or cloud servers—can’t operate on the battlefield. Defense AI must function even in deserts, in unstable networks, or with limited data. That’s why on-device AI capability is essential. Our strength lies in embedding software directly into hardware.”

Funzin’s expertise in communication continues to evolve. The company is now developing 6G hyperconnected network technology as part of a Defense Technology Promotion Research Institute project. The system will detect abnormal signals in advance, determine optimal networks in real time, and automatically switch frequencies in dynamic battlefield environments where multiple AI systems use different communication channels. Completion is targeted for 2026.

When asked whether handling such diverse technologies—from target recognition to recommendation systems, MLOps, on-device AI, and 6G networks—was challenging, Kim replied calmly,

“Whenever we need a solution, we develop it ourselves. That’s how our portfolio has grown.”

When asked what technological identity Funzin aims to maintain despite expanding into multiple fields, he answered,

“Robot intelligence.”

That remark explains why quotes from famous robot movies like Terminator, Transformers, and Star Wars are posted throughout the company’s offices. Funzin pivoted to the defense industry because, Kim said, “the military is the most open to robots.”

“Robotic weapons such as unmanned tanks and drones will become more common, and these robots will require advanced artificial intelligence,” he said. “We began developing AI models for robot intelligence because we already had automation tools, synthetic data platforms, and MLOps technologies to support them. In the private sector, investment in robotics is still limited. Like GPS—which was first developed for military use—robot intelligence will likely mature in defense before expanding to civilian applications.”

Vision for the Robot Brain

CEO Kim Deuk-hwa had a vision for robotic intelligence even before the advent of smartphones. After watching the Terminator series in high school, she dreamed of creating robots—but specifically focused on developing their “brains,” or intelligent systems, rather than the mechanical bodies themselves. She majored in radio engineering at university and later worked at SK Telecom, building expertise in mobile communication and vision AI. Since robots require networks capable of real-time sensor data exchange, communication infrastructure is essential for enabling AI “brains” that learn and adapt autonomously.

She joined Funzin in 2010 with the goal of transforming it into “a company preparing for robotic technology.”

“Funzin was founded by my husband, Kim Jeong-hoon, who specialized in mobile devices,” she recalled. “At that time, we often talked about AI robots—even before smartphones or the term AI existed. We imagined that adding communication capabilities to a portable media player (PMP) would eventually lead to a smartphone, and adding algorithms and intelligence would create artificial intelligence. Combining that with robotics became our long-term goal.”

When asked whether she had ethical concerns about developing AI that could be used in autonomous weapons, Kim replied,

“AI behaves only as it is taught by humans. The most important thing is social consensus on how to use technology.”

She emphasized the neutrality of technology while underscoring that defense AI is also a matter of national competitiveness.

“If our allies are all equipped with AI and robotic systems while we are not, we cannot remain a reliable partner. Some say autonomous weapons are necessary for national defense. Ultimately, we must be able to protect our citizens.”

At the end of the interview, Kim mentioned that her ultimate goal is to create companion robots—machines that care for the human mind.

“While I am proud of defense robots that make decisions and lead the military, I also want to develop care robots that comfort people,” she said. “As loneliness and anxiety increase, robots that provide emotional support will become more necessary. The AI embedded in such robots will represent the culmination of human knowledge and emotion accumulated over thousands of years.”

Ironically, Funzin plans to continue upgrading its military AI until it can eventually realize that vision.

https://www.forbeskorea.co.kr/news/articleView.html?idxno=400298