Research Overview
Data Digitalization
Much of the chemical and materials literature remains locked in legacy formats, limiting AI systems' ability to leverage this knowledge. Our research focuses on converting scientific literature into structured, machine-readable forms to enable large-scale aggregation and analysis. Key research foci include:
1) extracting data from scientific graphics,
2) reasoning over multimodal data and inferring missing details, and
3) establishing metrics to assess reproducibility and reusability of digitized data.
Digital Electrosynthesis
Organic electrosynthesis offers unique advantages such as enhanced selectivity, higher reactivity, and milder reaction conditions. Yet, reaction discovery still relies heavily on trial-and-error. Our group integrates data-driven and AI-enabled approaches to transform electrosynthesis research. Our efforts span the full research pipeline, from initial discovery and optimization to reactor design and scale-up. Key research areas include:
1) discovering reaction design heuristics at scale,
2) transferring insights from conventional organic synthesis to electrochemical reactions,
3) developing descriptors that capture electrosynthesis conditions for improved model performance; and
4) predicting optimal electrosynthesis reaction conditions.
Mechanism-Guided Self-Driving Laboratories
We develop chemistry-aware AI systems to guide rational, mechanism-driven reaction discovery and catalyst design beyond black-box predictions. Focusing on heterogeneous electrocatalysts, our research includes:
1) autonomous reasoning over multimodal experimental data to resolve reaction pathways and kinetics,
2) high-throughput synthesis of model heterogeneous electrocatalysts,
3) development of AI models that perform mechanistic reasoning to generate hypotheses and propose targeted experiments, and
4) development of self-driving laboratories that couple autonomous insight generation with iterative experimental design.