Hospital as Developer: Eirbot and Medical Sovereign AI

Switch

Homepage

facebook

Other Hospital CMUH APP icon

CMUH App

About CMUH

News and Award

:::

Home/

About CMUH/

News and Award/

Hospital as Developer: Eirbot and Medical Sovereign AI

About CMUH

2025/10/21

Taiwan's healthcare system is facing significant challenges due to a shortage of nurses. To tackle this issue, China Medical University Hospital (CMUH) has partnered with EverBot Technology to develop nursing robots that will assist frontline healthcare professionals. CMUH Superintendent Dr. Der-Yang Cho and EverBot Chairman Dr. Yu-Cheng Lee recently shared their insights on this initiative with IBMI.

CMUH's Dr. Der-Yang Cho (left) and EverBot's Dr. Yu-Cheng Lee (right) with the Eirbot nursing robot.

As dusk falls, the 20th floor of the medical building at CMUH in Taichung is illuminated. Nurses move quickly through the corridors, while "Eirbot," the nursing robot, glides along.

In June of last year, Dr. Chang-Hai Tsai, Chairman of China Medical University, asked CMUH’s Dr. Cho to evaluate the university's robotics initiatives. Dr. Cho recommended creating practical robots for quicker deployment, ultimately challenging himself with a question: “Who will benefit from these robots?”

"It's important to define the end user. If you make surgical robots, the users will be doctors. Transport robots can work for nurses, cleaners, and admin staff," Cho said. "What do hospitals need the most? Nurses!"

The positioning of Eirbot, described as “Designed for Nurses,” served as its foundational concept. Cho said that nursing is a complex profession, much like being a nanny, with numerous responsibilities. By reducing nurses’ workload, they can concentrate more on patients.

A Cross-Sector Push with Instant Chemistry

By chance, Cho met Mr. Khein Seng Pua, Chairman of Phison Electronics, and shared his robot concept. Pua introduced Phison's aiDAPTIV AI edge computing technology, and they both recognized the potential for integration.

Cho then considered Everfortune.AI, a spin-off company of CMU that specializes in medical AI software and has obtained several domestic and international certifications for medical devices. “We have medical knowledge and clinical expertise,” Cho stated. “It would be a good fit. The three of us hit it off immediately.”

The Eirbot was developed with remarkable speed. In April, Phison and Everfortune.AI established EverBot Technology. By mid-September, they launched the Eirbot prototype in just five months—an impressive feat in Taiwan's smart medical device industry, showcasing the team's dedication.

Eirbot's late-starter success stems from the shared philosophy of CMU, Everfortune.AI, and Phison: prioritizing practical applications. Cho stresses that the focus is on supporting nurses, rather than incorporating flashy features. CMU has guided the development from the start, ensuring that all functions of Eirbot have been validated by nursing staff, reflecting extensive clinical and therapeutic experiences.

Cho consistently challenges the development team with questions. Lee recalls, "When we presented sample dialogue scenarios, the Superintendent asked, 'If a user speaks and you take 10 seconds to respond, who would wait that long? And if your reply is thousands of words long, they’ll be long gone.' We made immediate adjustments."

After fine-tuning, Eirbot now responds to queries in 3 to 5 seconds, providing answers limited to around 80 Chinese characters, which match natural conversation flow and Cho's high standards.

AI “Brain” for a Broader Robot Portfolio

Lee explained that a robust backend system enables robots to share nursing duties. Eirbot serves as the conversational interface, while the real "brain" is in the backend AI servers, connected to 20 or 30 robots at once.

The server is integrated with hospital information systems, allowing the robot to recognize the patient's background, identity, and medical history during conversations. This integration optimizes efficiency and energy consumption, ensuring that each patient receives a suitable response.

Eirbot delivers patient health education.

EverBot will leverage Eirbot's experience to introduce multifunctional robots for cleaning, disinfection, item handling, long-term care, and home care services. "Robots are just one component of the smart healthcare service system. The entire data chain is ten to a hundred times larger," Lee explained.

On-Premises Servers for Security

When selecting a technical approach, Cho emphasized the use of on-premises servers instead of cloud services like ChatGPT. He acknowledged that while utilizing generative AI platforms is quick, it means handing over conversational content to third parties. “The accuracy and history of our data are valuable assets. Uploading our data is like doing their work for them. There is a risk.”

This collaboration with Phison reflects Cho's vision for “Medical Sovereign AI.” Cho explained, “We prefer on-premises servers because generative AI can produce hallucinations, and healthcare cannot afford even one error in ten million cases. Deploying on-premises ensures patient privacy by keeping personal data secure.”

Most importantly, all medical knowledge provided by Eirbot is validated by physicians. "CMU reviews every piece of Eirbot's data," Lee said. "To prevent errors, we do not include web-sourced information." This emphasis on data sovereignty and privacy distinguishes Eirbot from other AI robots.

Subscription Model

Following its launch, Eirbot will be promoted through a subscription model, with annual fees ranging from NT$500,000 to NT$800,000. This decision not only ensures long-term profitability but also symbolizes a commitment to service quality.

"Medical knowledge and technology advance rapidly. If Eirbot were sold under a buyout agreement, hospitals might find it impressive in the first year, but they could notice subpar performance by the second year and, by the third year, think, ‘How clumsy!’" Lee explained that a subscription model ensures ongoing communication, allowing for timely updates to backend servers based on evolving needs. "It's similar to how Apple sells iPhones. Software should be updated regularly."

Cho believes that this model establishes a commercial barrier. “As our system develops and new robots become more advanced, once we achieve PhD-level sophistication, competitors trying to catch up may find that they won't even see our taillights.”

Replacing 30% of Nursing Tasks

In functional planning, the R&D team adopts a pragmatic approach. Cho emphasizes that the goal isn't to replace nurses but to be a practical aid. “The first-phase target is to replace 30% of nursing tasks, with a long-term goal of reaching 50%.”

“The 1st-generation Eirbot offers four services: environmental navigation, health education, item delivery, and remote ward rounds. The 2nd generation will add contactless physiological monitoring and environmental surveillance. We will apply for medical device certification once the 2nd generation is launched.”

This phased approach stems from an understanding of industry realities. “We were clear about our capabilities,” said Lee. “This isn't about showing off. Medical robots must deliver reliability, privacy protection, and specialized knowledge bases. We won't overreach until these conditions are met.”

From Hospitals to Underserved Communities

Eirbot is currently in a three-month Proof of Concept (POC) trial at CMUH. “We plan to begin mass production once the POC concludes in December,” Lee said. “We will deploy one or two units on a single hospital floor and scale up.”

Eirbot's applications reach beyond hospitals. Cho presented his vision: “Remote indigenous communities could each have a unit serving residents at community health centers. Eirbot's video chat feature also allows doctors to conduct remote consultations. That’s telemedicine.”

EverBot team develops Eirbot to ease nursing workload.

In fact, his vision is materializing. “During a product demo, the pharmacy department requested one unit to be placed beside the medication pickup counter to address patient inquiries. The obstetrics and gynecology department is also enthusiastic. When nurses are on leave, there's no one to provide patients with health education. Having Eirbot will be a huge help.”

Many Taiwanese tech companies are united by a common goal: to leverage Taiwan's technological strengths to reduce nurses' workload and improve patient care. The experiences of CMU and EverBot show that smart healthcare can thrive when hospitals transition from being passive users to active developers. This understanding is crucial as healthcare industry players worldwide begin to look to Taiwan for inspiration.

Please find news source:Hospital as Developer: Eirbot and Medical Sovereign AI

回上頁