AI expert He Kaiming's homepage updates information about two new trainees --

Both Hu Keya and Li Zongyi are Chinese and are top students. Hu Keya is a PhD student while Li Zongyi is a postdoctoral researcher.

So far, among the six students recruited by He Kaiming since he joined MIT, five of them are Chinese.

And this time, the two new Chinese members have also shone throughout the entire process.

Submit 'Hu Keya'

Hu Keya graduated from Shanghai Jiao Tong University for her undergraduate studies.

During her high school years, she attended the prestigious Fujian Normal University High School Affiliated School.

In 2021, Hu Keya joined the renowned ACM class at Shanghai Jiao Tong University to study computer science.

According to the official account of Shanghai Jiao Tong University's Zhiyuan College, Hu Keya has been a member of the Brain-Computer Interface Laboratory (BCMI) at Shanghai Jiao Tong University since her junior year, under the guidance of Professor

During that period, she focused her research on AI for Science - hoping to combine AI with brain science and process raw EEG signals through self-supervised learning, in order to help patients with depression and other mental health issues.

After several months of refining, she completed the paper titled Contrastive Self-supervised EEG Representation Learning for Emotion Classification as the first author.

This achievement has been accepted by EMBC, the top international conference on biomedical computing, and she has been invited to the United States to present an oral report as a result.

Meanwhile, she co-authored a paper on improving self-supervised learning results, which was also successfully included in the top conference, Cognitive Science 2025.

During the summer vacation of her junior year, Hu Keya went to Cornell University for an internship. Under the guidance of Professor Kevin Ellis and doctoral student Hao Tang, she participated in a research project aimed at improving the efficiency o

The achievement was eventually accepted at the world's top machine learning conference, NeurIPS 2024, with Hu Keya as the second author.

After completing a project, Hu Keya teamed up with PhD student Wen-Ding Li to turn their attention to the then highly anticipated AGI open competition, ARC Prize 2024.

The concept of ARC was initially proposed by François Chollet to evaluate the learning and reasoning abilities of AI when facing new problems, rather than relying on memory or repeated training for specific tasks to achieve success.

ARC Prize 2024 is centered on this benchmark, requiring participants to submit algorithms that can solve never-seen task sets.

In other words, to win, the participating models must demonstrate human-level thinking abilities in scenarios with few samples and abstract reasoning.

The competition standards are rigorous, but the allure is equally astonishing with a total prize of over 1 million dollars, attracting 1430 teams from around the world.

To stand out in this challenging task, Hu Keya has led the research on a set of critical methods: using program synthesis to automatically generate data sets, fine-tuning large language models based on this dataset, and combining with test-time finet

This method has proven to significantly improve model performance.

After a fierce competition, the achievements of Hu Keya's team reached the state-of-the-art level at that time in the competition, winning the Best Paper Award.

Later, she co-authored an article summarizing the research findings and successfully published it at the top machine learning conference ICLR 2025.

Remember, at that time, Hu Keya was still just an undergraduate student.

During her undergraduate years, she has already authored four highly valuable papers, with half of them as the first author.

With such credentials, she naturally became a hot prospect for top universities when applying for a PhD. It is said that MIT, Princeton, Carnegie Mellon, Cornell, and the University of Washington all offered her a position.

Finally, she chose MIT for direct PhD admission.

Currently, Hu Keya is a first-year PhD student in the Department of Electrical Engineering and Computer Science at MIT, co-supervised by Kai-Ming He and Jacob Andreas.

She focuses on the intersection of language and vision research at MIT, aiming to create intelligent systems that can make more efficient use of data and have stronger generalization ability.

Another is the inventor of FNO, Li Zongyi

Li Zongyi, another disciple of He Kaiming who joined recently, has become somewhat famous in the AI academic circle before.

In 2021, Li Zongyi, who was still a doctoral student, published a heavyweight paper as the first author.

《Fourier Neural Operator for Parametric Partial Differential Equations》。

You might not be familiar with the name right away, but it was this paper that introduced the Fourier Neural Operator (FNO), which later became renowned, and first realized the true sense of large-scale application of 'neural operators'.

Before talking about FNO, we need to clarify a question: what is a neural operator?

In simple terms, a neural operator is a neural network that can learn to solve physical equations.

During training, scientists feed it with a large amount of data related to physical equation calculations, such as different starting water temperatures leading to the time needed to cook instant noodles; different release angles and speeds resulting

By learning from these samples, neural operators will eventually emerge with a set of universal rules, just like mastering a higher-level multiplication table.

In the future, when encountering similar data, it won't need to calculate step by step, but rely on intuition to instantly solve the final exam question.

For example:

If you want to predict where the storm will move, traditionally, even supercomputers have to struggle for hours to make a prediction;

However, if neural operators are used, with the input of the day's atmospheric pressure, temperature, and wind speed, it can calculate the path of the storm in just a few milliseconds.

It's like an unforgettable academic superstar who can glance at a question and immediately write the answer on the answer sheet with their vast knowledge, without the need to step-by-step calculate on paper.

（Cough cough, if the college entrance examination is done in this way, it will still be given a score of 0）

With this second-by-second capability, neural operators are highly effective in weather forecasting, carbon sequestration, and aerodynamic simulations.

Moreover, it enables AI to generalize at the level of physical laws for the first time, thus becoming a critical bridge connecting machine learning and fundamental sciences.

Li Zongyi's research focuses on this field, and they have taken a crucial step in this direction - their proposed FNO accelerates the operation of neural operators by moving the traditional spatial convolution into the frequency domain and using Four

This breakthrough has made FNO a milestone model in the field of AI for Science, and Li Zongyi has been recognized as one of the core contributors in the direction of neural operators, with his Google Academic citations exceeding 12,000.

Li Zongyi is currently a postdoctoral researcher at MIT under the guidance of Professor Kai-Ming He.

But his stay at MIT may not be long

It is reported that he has obtained the position of assistant professor at New York University and will start working in the fall of next year.

Li Zongyi comes from Beijing. He once studied at Renmin High School and then went to the United States for further education.

During his undergraduate studies, he pursued a double degree in Computer Science and Mathematics at Washington University in St. Louis, while also minoring in Jazz music.

In 2019, he went to California Institute of Technology to pursue his PhD under the guidance of Anima Anandkumar and Andrew Stuart, who both witnessed the birth of FNO.

In addition, during his PhD studies, Li Zongyi also interned at NVIDIA for three consecutive summers.

What is He Kaiming doing? Deep Dive into AI for Science

In fact, at his job interview speech at MIT in 2023, Assistant Professor He Kaiming made a statement about his focus on AI for Science in the coming years as a key area of exploration and research.

Nowadays, if we look at his team composition, it indeed has profound significance.

The newly joined members, Hu Keya and Li Zongyi, one who has done research in the Brain-Computer Interface Lab at Shanghai Jiao Tong University, and the other who is a leader in the field of neural operators, are just the right match in this directio

With previous mentors Deng Mingyang, Bai Xingjian, Li Tianhong and Jake Austin, He Kaiming has amassed six outstanding students, forming an impressive lineup.

However, Quantum Bit heard that the original plan for this year was actually seven people.

The last candidate is a male with an equally impressive resume. There are even rumors that his recommendation letter was written by a top professor on the level of He Kaiming, and he is highly recommended.

However, he is not yet featured in the introduction of He Kaiming's team.

As one of the inventors of ResNet, Kaiming He has benefited more outstanding young people since his departure from Meta to join academia in 2024. This has given young scholars the opportunity to further advance AI fundamental research and make pionee

In fact, the four members of the ResNet team are all cultivating the new generation in their own ways.

Kaiming He's main profession is as a MIT professor, but he also has industrial collaborations with Google DeepMind.

Zhang Xiangyu's primary position is as the chief scientist of Jueyuemingxing, but he also holds a part-time position as a professor at his alma mater, Xi'an Jiaotong University.

Ren Shaoqing's primary occupation is as the vice president of NIO in charge of driving autonomy. However, this year he has also served as a professor at the artificial intelligence laboratory of his alma mater, the University of Science and Technolog

Believe that the new generation of AI talents is among those they are teaching and helping to pass on knowledge to.n

Under the guidance of master teachers, students grow steadily and ceaselessly.

Reference link:

[1]https://lillian039.github.io/

[2]https://zongyi-li.github.io/assets/pdf/Zongyi_CV_July2025.pdf

[3]https://www.facebook.com/lizongyijohnny/?locale=zh_CN

[4]https://zongyi-li.github.io/

[5]https://mp.weixin.qq.com/s/xTXjyE3MFLHGpVpJEXIrYA

This article is from the WeChat public account "Quantum Bit"