Zijing Zhikang Releases AI-Native Diagnostic and Treatment Operating System, with 54 AI Doctors Teaming Up to Collaborate

Currently, global healthcare systems face the core challenge of a shortage and uneven distribution of high-quality medical resources.

As population aging accelerates, the rigid demand for chronic disease management and health follow-up has surged, and the traditional medical supply model can no longer carry the demand for high-quality medical services.

"The vast majority of currently approved AI medical products are auxiliary tools with fixed functions, focusing on single-point tasks such as lung nodule CT detection, fundus screening, medical record generation, and imaging triage. They are embedded in existing workflows and ultimately confirmed by licensed physicians; their capabilities are locked at the factory and do not automatically become stronger because they have 'seen more patients,'" said a medical expert.

A developer of AI medical tools admitted frankly: "Insufficient real-world clinical data and the failure of medical institutions to share information greatly constrain AI training. Because real diagnostic and treatment data are scattered across different hospitals and different systems, with inconsistent formats and varying standards, cross-institutional aggregation faces privacy compliance and interest coordination difficulties, forming 'information silos.'"

This directly leads to training data being skewed in distribution across diseases, populations, and institutions, making it difficult for models to cover the diversity of real clinical practice, and rare diseases and long-tail scenarios have long lacked sufficient samples. Facing the massive follow-up and chronic disease management needs brought by an aging society, this "passive query" tool model clearly cannot meet the systemic requirements of the future national layout of AI medical institutions—it can neither achieve real-time cross-specialty collaboration nor support the continuous health management of primary care.

On October 21, Zijing Zhikang, incubated by the Institute for AI Industry Research (AIR) at Tsinghua University, released the "AI-Native Diagnosis and Treatment Operating System." This system allows AI doctors to "team up" and collaborate like an expert team, with doctors making the final decisions, and each side learning while being used, increasing efficiency by 30%.

"The core contradiction in global healthcare is the insufficient supply of medical resources and their unbalanced distribution. This is the starting point of all our medical innovation," Qiao Yuchen said, citing data that by 2030 the global shortage of medical personnel is expected to reach 11 million, and the gap in per capita medical spending between high-income and low-income countries will reach 875 times.

Zijing Zhikang's "AI-Native Diagnosis and Treatment Operating System" has built 54 AI doctors and millions of hyper-realistic AI patients, covering more than 20 clinical departments and tens of thousands of diseases. The AI doctors' test accuracy on the MedQA dataset exceeds 96%, far surpassing the average human expert score of 87%.

More noteworthy is its "dual-layer evolution architecture": AI doctors first receive high-throughput training in a virtual world composed of massive numbers of AI patients, completing tens of thousands of consultations in a day; after reaching a certain level, they are deployed to physical hospitals to form a companion collaboration relationship with human doctors and continue to evolve in real diagnosis and treatment.

The technical source of this foundation comes from the evolvable agent technology of Professor Liu Yang's team at Tsinghua University. The internationally authoritative academic journal Nature has published an article reporting Agent Hospital (Zijing AI Hospital), built by the Tsinghua University team, as a core case of Chinese AI Agents being implemented in the medical field.

At the launch event that day, the president of Riphah International University in Pakistan announced via video that the two sides have established a strategic-level cooperative relationship and hope Agent Hospital will help it build Pakistan's first artificial intelligence hospital.

"An agent will never see anything beyond the knowledge it understands, but a doctor needs to see the whole picture and then make the most correct order of handling," said an emergency department physician.

Another neurosurgeon values the change in the preoperative discussion session: "The Agent participates in recording and summarizing the discussion content, which can basically save about half the workload while reducing omissions."

"The emergence of agents will greatly shorten and lower doctors' learning curves and practice costs, and supply-side problems can begin to be addressed relatively well. We need to find the point that achieves maximum effect with minimal effort, rather than applying force evenly," said Lu Gang, a partner at Legend Star.

"AI is a tool; it can amplify doctors' capabilities rather than replace doctors. AI solves the problem of information processing, while doctors solve the problem of human disease. The best state is for AI to handle complicated affairs while doctors focus on treating patients, jointly serving patients," said a clinician.