Tairex, an AI healthcare company incubated by Tsinghua University's Institute for AI Industry Research (AIR), has launched a framework it calls Multidisciplinary Human-Agent Collaboration, or MHAC, and upgraded its Agent Hospital system. The company presented it at an event at the China National Convention Center in Beijing on October 21.
Agent Hospital itself is not new. Its research origins lie with AIR professor Liu Yang's team, which built a simulated hospital where AI doctor agents treat millions of synthetic patients, reflect on failures and distill reusable rules. What is new is the step from single AI assistants to coordinated avatars of different specialists working inside ordinary outpatient visits.
The target is the multidisciplinary team consultation, in which experts from several departments review one complex case. Tairex argues such consultations come late, after patients have visited several departments, are limited to large hospitals and offline meetings, and are expensive to convene. MHAC tries to bring that input into routine clinics.
In practice, a doctor sees patients as usual. As new information arrives, the system detects when another specialty is relevant and shows that specialist avatar's advice as a prompt. In one example, a patient seen in endocrinology for hyperthyroidism also has a history of asthma, so a respiratory avatar flags medication risks. For cases that clearly need several specialties, multiple avatars join a consultation view with suggestions on diagnosis and drugs. Doctors can see which avatar each suggestion came from, and the human doctor makes the final judgment.
Avatars are meant to learn from the real doctors they represent. A company executive gave one example: the AI recommended the most effective drug, but the doctor chose a cheaper one covered by medical insurance, and such corrections are fed back into fine-tuning. Tairex says that, with approval from both hospitals, avatars of outside experts could join a consultation.
According to the company, as of July the system included 54 AI doctors and millions of simulated patients across thousands of diseases, and its virtual clinic covers 27 departments. At Beijing Tsinghua Changgung Hospital, 18 clinical departments have run outpatient tests with more than 40 doctors using it regularly. The company and hospital report efficiency gains of about 30% and MedQA accuracy above 96%. These figures have not been independently verified, and the AI does not make diagnoses on its own.
SEE ALSO: BIGAI Spinout Tongyan Lingzhi Bets on Personal LoRA