Trustworthy AI, scientific discovery, and graph foundation models.
I work on robust and adaptive learning over structured data, with a particular interest in graph neural networks, probabilistic neighborhood selection, and reliable machine learning systems for scientific settings.
I am a researcher working at the intersection of graph machine learning, trustworthy artificial intelligence, and structured data modeling. My recent work focuses on adaptive neighborhood sampling, reinforcement learning driven graph structure optimization, robust representation learning, probabilistic neighbor selection, and graph purification under noisy and heterogeneous settings.
More broadly, I am interested in developing principled and scalable learning frameworks for graph structured data, with applications in scientific discovery, recommendation systems, knowledge enhanced reasoning, and reliable decision making. My long term goal is to build a unified theory and system for adaptive information acquisition in large scale graph learning.
All publication and achievement information on this page is fictional and used only for front-end design practice.
Advisor: Prof. Alex Thornton. Research on graph foundation models, robust graph representation learning, and adaptive information acquisition in structured domains.
Thesis focused on reinforcement learning driven adaptive neighborhood sampling and dual similarity guided neighbor selection in graph neural networks.
Built strong foundations in mathematics, optimization, machine learning, scientific computing, and algorithmic modeling.
Built an end to end research infrastructure for training, evaluating, visualizing, and benchmarking adaptive graph learning algorithms across citation networks, recommendation data, biological interaction graphs, and large scale knowledge graphs.
Developed graph foundation model based pipelines for molecular property prediction and interaction mechanism discovery, integrating domain priors, uncertainty estimation, and multi-view graph representations.
Designed graph purification and influence aware propagation techniques to improve reliable classification and anomaly detection in noisy graph environments.
Explored dynamic graph construction and adaptive neighborhood retrieval for improving reasoning quality in graph augmented large language models.
Teaching Assistant, MIT, Fall 2028
Guest Lecturer, Spring 2029
A modular toolkit for adaptive sampling, graph purification, and robust graph representation learning.
An open benchmark for noisy and heterophilous graph learning with reproducible evaluation protocols.
Python, PyTorch, PyTorch Geometric, Deep Learning, Scientific Computing, Large Scale Experimentation, Web-based Research Visualization
Email: jinyh23@163.com
GitHub: https://github.com/KaimingHe
Google Scholar: https://scholar.google.com/citations?user=DhtAFkwAAAAJ&hl=en/
Office: Stata Center, MIT, Cambridge, Massachusetts, USA
Research Collaboration: I am always open to collaboration on graph learning, trustworthy AI, scientific machine learning, and adaptive decision making systems.
Interested in collaborations on graph learning, robust AI systems, scientific discovery, and adaptive reasoning over structured knowledge.
Start a ConversationA few quiet dreams I hope to carry into the years ahead.
Beyond research and equations, I carry a quiet yet persistent curiosity about the world itself. There are places I long to reach not for achievement, but for experience, to stand in the silent snow of Hemu in Xinjiang, to feel the rhythm of the road while cycling along the eastern and western coasts of Taiwan, to walk through the streets of New York where ideas and cultures converge, and to wander across the vast landscapes of Australia under an endless sky.
These are not merely destinations, but fragments of a larger dream, to witness the beauty of this world with my own eyes, to understand different lives, and to let each journey reshape how I think, feel, and create.
And perhaps, somewhere along this journey, it would mean even more to share these moments with someone who sees the world in a similar light, a companion not only in distance but also in thought, someone to walk beside me through unfamiliar cities, to ride across long horizons, to stand quietly in vast landscapes, and to turn fleeting moments into something lasting.
In the end, I believe that research defines how far we can think, while exploration defines how deeply we can live, and the most meaningful journeys are those that are shared.