Research

My research lies at the intersection of neural and symbolic AI, integrating deep learning and foundation models with symbolic knowledge and structures including knowledge graphs and ontologies. I develop hybrid AI systems that are not only powerful in prediction but also reliable, interpretable, and generalizable, ensuring that they can be trusted in high-stakes domains such as biomedicine and healthcare.

To achieve this, I conduct fundamental research in the following areas:

Knowledge-enhanced LLMs/foundation models

Integrating knowledge with large language models to improve their reliability, interpretability, factual accuracy, or generalization ability
Representative work: ICLR’25, EMNLP’25, NAACL’25, EMNLP’24, NeSy’25

Foundation models for structured data & knowledge – Developing generalizeble models for structured data
Representative work: EMNLP’25, ICML’25

Representation learning with data & knowledge structures

Learning on complex graph structures — Developing methods for learning from graph-structured data, including knowledge graphs and ontologies.
Representative work: ICML’25, WWW’25, NeurIPS’22a, ACL’23, AAAI’24, CIKM’24, KDD’22

Semantic web and ontologies — Working with description logic and ontology reasoning for structured knowledge representation.
Representative work: ISWC’22, ISWC’23

Geometric representation learning — Exploiting data geometry (hyperbolic, pseudo-Riemannian spaces) to improve machine learning on hierarchical and complex-structured data.
Representative work: NeurIPS’22a, NeurIPS’22b, KDD’22

Trustworthy AI in healthcare and biomedicine

Neuro-symbolic machine learning — Combining neural networks with symbolic reasoning by imposing structure and prior knowledge in machine learning models.
Representative work: NeurIPS’22b, ISWC’22, ICDE’24

AI for healthcare and biomedicine — Applying machine learning techniques to medical and healthcare domains.
Representative work: ICML’25, AMIA’24, SIGIR’23