Singaporean roboticist behind AI model that outperformed ChatGPT leaves US to join Asia's top university
Singaporean robotics expert Jiafei Duan, who helped develop an AI model that performed better than OpenAI's GPT-5 in special tests, has returned home to work at the National University of Singapore (NUS).
Duan joined NUS as a Presidential Young Professor in July. He had completed his doctorate at the University of Washington in the U.S. in May, according to an NUS announcement on Oct. 19. NUS ranks 10th in the world in the QS World University Rankings 2027, making it the highest-ranked university in Asia.
Before returning to Singapore, Duan led the development of MolmoAct2, an AI model designed to help robots understand what is around them and carry out physical tasks. He worked with researchers at the Allen Institute for AI and the University of Washington in the U.S. to develop it. According to the research team's findings, the model's system for working through problems received the highest average score across 13 robotics-related tests. It performed better than GPT-5 and Google's Gemini Robotics-ER 1.5.
Duan previously worked as a research intern at Nvidia, a major U.S. computer chip company. He has also received recognition for his work. In June, he was named a finalist for a Rising Star Award in spatial intelligence, which is about understanding objects and the space around them. The award was part of a workshop connected to CVPR 2026, an international conference about computer vision, or how computers understand visual information.
Singaporean roboticist Jiafei Duan. Photo courtesy of NUS |
Duan became interested in robots when he was a teenager. At age 16, he built a robot that could move using PVC pipes. He later experimented with electronics and technology that tracks movement, he said in a July interview with the 22Astronauts podcast.
While studying at Singapore's Ngee Ann Polytechnic, Duan and another student developed a robot that could be controlled by the mind. They got the idea from the Jedi characters in "Star Wars." The robot was designed to help dispose of bombs. Their invention won the Tan Kah Kee Young Inventors' Award.
Duan later graduated from Nanyang Technological University with the highest distinction in electrical and electronic engineering. He then went to the University of Washington to do research for his doctorate, where he worked under professors Dieter Fox and Ranjay Krishna, according to his personal website.
Since returning to Singapore, Duan has received S$2 million (US$1.6 million) in funding through the country's National Robotics Programme. The money will help him develop technology that allows robots to carry out everyday tasks outside carefully controlled laboratories.
His project involves collecting information from robots working in real-world places and creating shared ways to measure how well they perform. Duan told NUS, "The goal is to help robots learn outside carefully controlled laboratories and collect information about the many different and complicated tasks that robots will actually face in the real world."
His team is working with FrodoBots, a robotics data company that has collected information from human-shaped robots working in homes across Southeast Asia. The research could eventually help robots carry out laboratory experiments, assist with manufacturing and do household tasks.
Explaining why he decided to return after studying and doing research in the U.S., Duan said, "I wanted to come home and help Singapore's growing community working on AI that can understand and act in the physical world, as well as robotics. I especially hope to help build the research community here and train the next generation of researchers and robotics experts."
Over the next five to 10 years, Duan hopes to develop robots that can help older adults at home and handle physically demanding work in industries. He wants to measure their progress by how much these robots improve people's quality of life.