My name is Bo Li. I am currently a postgraduate student at NUS with a background in Materials Science.

My research interests lie at the intersection of artificial intelligence and computational chemistry, with a focus on data-driven strategies to accelerate materials discovery.

My prior work includes inverse materials design using transformer-based architectures and the development of a novel adapter module for classifier-free guidance. I have also explored the use of Large Language Models (LLMs) in multi-agent AI systems to enable text-guided molecule discovery.

In the future, I aim to establish an automated materials discovery pipeline encompassing data mining, generative models for novel materials, property emulators that integrate machine learning with quantum mechanical methods, and domain-specific LLMs to execute and optimize experiments within a closed-loop discovery framework.

🎓 Education

Research perspective

Interesting topics

1. Recently, I read a short report questioning whether inverse design truly accelerates materials discovery. Even advanced models like MatterGen sometimes suggest novel crystalline materials that, when synthesized, turn out to already exist — but in amorphous rather than crystalline form. This raises an important question: can we propose a new graph representation for amorphous materials? see this article

2. The next major challenge is often described as the “elephant in the lab”: how can we connect theoretically accessible materials to practical synthesis routes? Generative models suggest materials that are challenging or even impossible to synthesize in reality. For further discussion, see this article on ChemRxiv.

Above are issues that I want to solve, but automatic pipeline is my final aim. Below is my desired workflow for my future academic journey !