Developing generative AI approaches to create novel drug-like chemical structures with targeted properties. Research will focus on exploring and exploiting the chemical space efficiently to identify molecules that are both innovative and synthetically feasible.

Exploring, protoyping, and testing interpretable ML frameworks that provide transparent, scientifically grounded explanations for model predictions. Research will include understanding how generative and predictive molecular models understand feature patterns, uncovering relationships between learnt molecular representations and actual physical features to make AI-driven drug design and discovery more understandable, explainable, and trustworthy.

Building and evaluating predictive models to estimate key drug properties such as solubility, toxicity, and binding affinity at early stages of molecular design. The goal here is to guide candidate selection more effectively and reduce reliance on costly experimental trial-and-error.

Designing methods to learn rich, multi-modal representations of molecules from diverse data sources, including structural, physicochemical, and biological information. These representations aim to capture the underlying features that drive molecular behavior, supporting downstream prediction and generation tasks.

Developing models that treat chemical structures as language objects, enabling the application of natural language processing techniques to molecular representations like SMILES, SELFIES, and molecular graphs. Research focuses on learning chemical syntax and semantics to build foundational models that can learn robust chemical representations, generate novel molecules, and capture structure–property relationships.

Building computational pipelines and machine learning frameworks for analyzing, managing, and predicting molecular properties. Work involves integrating large chemical datasets with AI models to accelerate hypothesis generation, virtual screening, and decision-making in drug discovery.