Fenrir Data Analytics: Veteran Self-Taught in AI Aims to Be "Taiwan's Palantir"
The AI model DeepSeek, developed in China, was released at the end of January 2025, shocking the world and sparking discussions about cybersecurity issues. The U.S. military, the Japanese government, and the Premier of Taiwan, Cho Jung-tai, all issued bans on DeepSeek.
At the heart of the issue is the concept of "sovereign AI." Sovereign AI refers to AI models developed by governments or citizens of different regions to protect their own political, economic, and cultural interests. In other words, sovereign AI is linked to security and national defense. One of the key questions raised by Fei Ching-Hua, CEO of the startup Fenrir Data Analytics, is whether Taiwan can have its own AI that it can use and trust.
Fei Ching-Hua recalls, "During Nancy Pelosi’s visit to Taiwan, I was glued to the screen, constantly watching for any changes that might happen. I couldn’t sleep, and my palms were sweating the whole time. Every little movement sent a jolt through my nerves." He was referring to the situation in 2022 when Pelosi, the Speaker of the U.S. House of Representatives, visited Taiwan, prompting China to conduct days of military drills around the island. At the time, Fei Ching-Hua, who was still in the military, was deeply affected by the experience.
"We need more technology to help people determine the best immediate response strategies, but this is different from buying military weapons. This time (with AI), we cannot rely on outsiders," he emphasized. Starting with the defense sector, building an AI that Taiwanese people can trust became his mission.
How to overcome doubts about military veterans starting businesses after injury?
At the end of 2021, Fei Ching-Hua was injured during military training and was forced to transfer to the information department. "I was on the front lines with many colleagues, so I knew the biggest issue facing Taiwan's military is the lack of talent," Fei Ching-Hua said. He believes that Taiwan's defense still relies on human experience and hardware thinking to assess situations. However, with gaps in manpower and military operations heavily relying on soft skills like command and control, he sees a future where software technology will be essential in formulating strategies. And AI is the answer.
"Actually, what we want to do is somewhat like the Taiwan version of Palantir," Fei Ching-Hua said with a smile. Palantir is known for its big data analysis and AI, with major clients including the U.S. Department of Defense and military applications, a vision that aligns perfectly with the direction Fei Ching-Hua's company, Fenrir Data Analytics, intends to take.
Having ideals, passion, and goals doesn't necessarily mean that a concept will succeed. On his entrepreneurial journey, Fei Ching-Hua encountered two major obstacles: technical capability and the mindset needed to face skepticism.
"When a retired soldier starts a tech company, people will naturally question our abilities. Some might even say, 'Did you get the contract because of connections or collusion?' All we can do is keep proving our strength," Fei Ching-Hua said.
From his injury and discharge to founding his company in early 2023, Fei Ching-Hua taught himself programming and AI, while also recruiting former partners from the military’s information department to join his team. Fei Ching-Hua admitted that while his team might not always beat engineers with years of experience in technical skills, what matters most in the defense sector is not just the AI's capabilities, but its credibility.
Explanation and data are the two key elements of defense AI.
"AI experts don’t understand military operations, and military experts don’t understand AI. These two groups have never intersected," said Fei Ching-Hua. This is one reason he found it difficult to initiate innovative R&D within the military. Military work requires rigor, discipline, and rules, so entrepreneurship became his way to quickly find innovative solutions. Therefore, Fenrir Data Analytics' goal is to create an AI product tailored to military habits.
Fei Ching-Hua explained that the key lies in "explainability." Every decision must have a reliable basis, not just a black-box answer. To achieve this, Fenrir Data Analytics developed the "TWIN Framework" AI model architecture, using a modular approach to ensure that every decision made by the AI can be traced, and based on this framework, they developed the military defense product, the "Taiwan SHIELDs System."
The "Taiwan SHIELDs System" is primarily used for simulating military drills. By analyzing documents (LLM), GIS data, and other information, the AI learns about physical factors such as road width, weapon load, and vehicle fuel consumption. It then creates a digital twin system for drills, helping the military plan training frameworks and strategies for different scenarios.
For example, in a scenario where military training involves responding to a "cut underwater cable," the AI would analyze whether it's more efficient to clear the sea area immediately or issue a warning. If the decision is to clear the sea, the system can predict the sea area boundaries, vessel inspections, required time, potential casualties, and more, enabling the military to quickly devise the best response strategy.
Not just defense! Fenrir Data Analytics plans to bring decision science to civilians.
However, Fei Ching-Hua admitted that the product is still in the development phase and will need to be used in the military to reach perfection. The "Taiwan SHIELDs System" is currently undergoing a PoC at the National Chung-Shan Institute of Science and Technology. If the system is successfully validated and implemented, the revenue model will be based on a buyout with ongoing maintenance. "Typically, a military procurement deal will exceed one billion New Taiwan Dollars (approximately 32 million USD), and we hope to reach that point," Fei Ching-Hua said.
However, product development requires significant funding and talent. As a result, Fenrir Data Analytics has also developed an educational version of its product, the "International Crisis Decision-Making Training System." Through gamification, this system trains new recruits in decision-making. It not only helps maintain revenue but also alleviates the defense crisis caused by a shortage of talent. The product has already been implemented at the National Defense University.
"In fact, I don’t want to limit myself to the 'defense sector.' We are more of a decision science team. Therefore, we look forward to applying our simulation software to the civilian sector in the future, such as disaster response, energy management, and more, because these areas also require offensive and defensive drills," said Fei Ching-Hua.
