Associate Professor and Doctoral Supervisor at the National Engineering Laboratory for Big Data, Shenzhen University (a recipient of the "Hundred Talents Program").
Background:
2012-2016: Admitted to Peking University's Department of Mathematics through an undergraduate mathematics competition.
2016-2021: Pursued a PhD in Computer Science at Peking University, receiving the Apple PhD Scholarship (only one recipient in mainland China).
Graduated with awards including Outstanding Graduate of Beijing, Outstanding Graduate of Peking University, Huawei Genius Youth, and Tencent Technical Expert.
Received honors from ACM SIGSAC and the China Electronics Education Society.
Recruitment within the Research Group:
Research topics and directions will be customized based on students' individual interests. Academically, the focus is on publishing papers in Nature sub-journals and CCF A-level papers. In terms of practical applications, students are encouraged to explore new commercial product forms in the era of large-scale models, and assistance will be provided in connecting students with relevant corporate resources for implementation.
1. We are continuously recruiting undergraduate students, students applying for postgraduate studies, or students preparing for postgraduate entrance exams who are interested in my research are welcome to apply for internships. Undergraduate students are encouraged to join our group during the summer of their sophomore year, with an internship period of no less than 6 months.
2. We have abundant computing resources, with over 20 local 4090 computing power servers and sufficient H2O cloud computing power.
We welcome students who are committed to (1) exploring cutting-edge model alignment technologies and (2) building entirely new AI-native commercial products in the AGI era to join our research group! Please send your CV, research plan, and supporting materials in a single email package to wubingzheagent@gmail.com & wubingzhe@szu.edu.cn;
Research Directions
1. Large Model Alignment and Security
1) Construct secure long-inference & slow-thinking models and apply them to robotics, finance, and other scenarios.
2) Construct trusted tool calls and RAG systems; research security attack and defense methods.
2. Applications
1) Finance: Construct a multi-agent risk early warning system for the financial market and apply it to downstream financial risk control, secondary market risk monitoring, and other fields.
2) Biomedicine: Design out-of-distribution optimization methods to address model reliability issues in downstream applications such as two-photon images, pathological images, and genomics.
3) Robotics: Construct corresponding alignment technologies for edge-side trusted embodied intelligent models.
Basic Requirements for Students:
1. Proficient in using various AI tools, including Cursor, Coze, LangChain, etc.
2. Solid foundation in machine learning and mathematics.
3. Strong programming skills and system optimization ability are a plus.
4. Possesses self-motivation and a willingness to explore cutting-edge technologies and products in the upcoming AGI industry cycle;
Personal Profile
Bachelor's and Doctoral degrees are both from Peking University, under the supervision of Professor Sun Guangyu. The applicant has long been dedicated to research in AI security and trustworthy AI, publishing over 40 papers in top international journals and academic conferences, including:
(1) 20 CCF (China Computer Federation) Class A conference or journal papers;
(2) In the past five years, the applicant has published 13 CCF-A conference or journal papers as the first author or corresponding author, including 1 ICML oral paper (acceptance rate 2.3%) and 2 NeurIPS spotlight papers (acceptance rate 5%);
(3) According to Google Scholar, the applicant's papers have been cited over 1900 times in the past five years.
(4) Has received awards for Outstanding Doctoral Dissertation from ACM SIGSAC and the China Electronic Education Society.
In addition, the applicant has extensive experience in the industrial transformation of academic achievements. The applicant is committed to applying the above-mentioned achievements of trustworthy AI to different interdisciplinary fields: (1) Open domain risk governance: Based on the basic model, a more intelligent risk governance agent is built for a series of scenarios including content review, financial risk control, and text-to-image bias governance. Some of the above research results were applied to the Tencent Charity risk control scenario to combat black and gray industries and won the Tencent Sustainable Social Value Award. (2) Life sciences: A series of out-of-distribution robust optimization algorithms are applied to drug discovery, pathological analysis, omics analysis and other scenarios. The cross-domain collaborative paper was published in Nature Methods with an impact factor of 58.
