Chinese Scholar Pioneers AI-Driven Life Sciences Research

When Xie Weidi, a 37-year-old associate professor at Shanghai Jiao Tong University (SJTU), decided to pivot his research from computer vision to the intersection of artificial intelligence and life sciences, he became what he calls a “rare species”—a researcher willing to bridge two of the most complex fields in modern science. That decision has already yielded remarkable results: DeepRare, the world’s first medical evidence-based reasoning agent for rare disease diagnosis, published in Nature in February 2026.

From Oxford to Shanghai: A Journey of Strategic Pivots

Xie’s path to the frontier of AI-driven life sciences was anything but linear. After graduating from Beijing University of Posts and Telecommunications in 2012, he pursued a master’s in computer vision at University College London, followed by a PhD at Oxford University’s renowned Visual Geometry Group (VGG). There, he studied under Andrew Zisserman, a pioneer in computer vision, and Alison Noble, a leading authority in medical imaging.

His time at Oxford was marked by an unusual challenge: for three years, his advisors refused to let him publish. “They couldn’t tolerate doing work that was ‘good enough’ and then sending it out,” Xie recalled in an interview with The Paper. “You had to do your best.” Those years of intensive reading—accumulating stacks of papers over a meter high—gave him a foundational breadth that would prove invaluable. In 2018, he published seven papers across computer vision and medical imaging, graduating with a PhD and staying on for postdoctoral research.

In 2021, Xie made another pivotal decision: he turned down a faculty position at Oxford to return to China. “Especially looking at it now, in terms of attention and investment in AI, our country’s support is enormous and the pace of development is very fast,” he told The Paper. He joined SJTU’s School of Artificial Intelligence in January 2022, and by 2023, he had shifted his focus from computer vision to AI for Science (AI4S), specifically life sciences.

DeepRare: A Paradigm Shift in Medical AI

The centerpiece of Xie’s work is DeepRare, an AI system co-developed with SJTU colleagues Professor Zhang Ya and clinical partners at Xinhua Hospital. Unlike traditional “black box” medical AI systems that provide diagnoses without explanation, DeepRare uses a “central hub + specialized agent” architecture that mimics a physician’s diagnostic process: observe symptoms, search knowledge bases, form hypotheses, and verify with evidence.

According to Jiefang Daily, DeepRare achieved 57.18% top-1 accuracy on phenotype-only diagnosis—23.79 percentage points above the previous best—and 70.6% accuracy when incorporating genetic data. The system’s diagnostic reports earned a 95.4% approval rate from expert physicians at Xinhua Hospital.

“Birth, aging, sickness, and death are the most important things for everyone,” Xie said, explaining his motivation. “If there is one field where AI can benefit humanity, it must be life sciences.”

The impact has been immediate. Since its online platform launched in July 2025, DeepRare has diagnosed over 15,000 cases globally, with more than 2,000 professional users from over 1,000 medical and research institutions registered on the platform. SJTU News reported that at the system’s official launch event in February 2026, SJTU also inaugurated the “AI & Biomedicine Innovation Center” and launched a “Global 10,000-Person Clinical Verification Initiative.”

The “Rare Species” Advantage

Xie’s career trajectory illustrates a strategic insight: researchers who bridge established fields with emerging frontiers become uniquely valuable. His background in both computer vision (from Oxford VGG) and medical imaging (from Alison Noble) positioned him perfectly for the AI4S transition.

“This kind of transition—if you say it’s hard, it’s because there’s a lot of domain-specific knowledge,” Xie explained in an interview with HyperAI. “But if you say it’s not hard, it’s because you become a ‘rare species’ willing to use AI for life sciences. People in life sciences can’t wait to discuss with you every day how AI can solve the problems they’re interested in.”

This interdisciplinary approach is increasingly vital. Rare diseases affect an estimated 20 million people in China across over 7,000 known disease types. Most grassroots doctors lack experience diagnosing these conditions, leading to years-long “diagnostic odysseys” for patients. DeepRare addresses this by functioning as a “digital quality controller” that helps doctors identify diagnostic gaps.

Building the Future: Guanyi Intelligence and Foundational Models

Xie’s ambitions extend well beyond rare disease diagnosis. In early 2026, he became CEO of Guanyi Intelligence, an AI life sciences company spun off from SJTU research. “What we want to do is build foundational models for the life sciences,” he told The Paper. “Using powerful AI model capabilities to help achieve the most challenging tasks—understanding disease mechanisms, designing new drugs, advanced therapies—and ultimately, of course, our dream is to cure all diseases.”

SJTU’s “three-in-one” innovation ecosystem—combining the AI School for research, the Shanghai Computing Court for engineering, and the Industrial Technology Research Institute for entrepreneurship—has been instrumental in this transition, creating a complete pipeline from original innovation to real-world application.

A Broader Trend: China’s AI Talent Pipeline

Xie’s story exemplifies a broader pattern: Chinese-trained undergraduates who gain elite international training and return to lead cutting-edge research. His receipt of the first Oxford-Google DeepMind full scholarship—worth approximately 1 million RMB—as the only Chinese recipient that year, and his subsequent return to China, reflect the “brain circulation” strengthening China’s AI research ecosystem.

As Xie told his students, “Artificial intelligence is a young endeavor—it’s the career of young people. Its development speed is very fast. We all need to keep learning, to see what the world’s cutting-edge AI technologies and models have achieved.”

What’s Next

With DeepRare already deployed in clinical settings and a global verification initiative underway, the team is now focused on scaling the system from 15,000 cases to over 20,000 clinical validation cases. The newly established AI & Biomedicine Innovation Center will coordinate across SJTU’s various schools and hospitals, while Guanyi Intelligence works toward foundational models that could transform drug discovery and advanced therapeutics.

The question Xie posed in his HyperAI interview remains central: “Defining the problem is more important than solving it. As long as a meaningful problem is defined, countless people will follow up and solve it.” For Xie Weidi and his team, the problem is clear—and the solution is just beginning to take shape.