• designed a BERT model-based recommendation system for electrolyte search over 10, 000 electrolytes, where a two-tower model was employed for candidate generation and Multimodal Mixtures of Experts (MMoE) were implemented for fine-ranking• developed an active learning framework for rapid molecule discovery over 2.8 million candidates, where a GNN-based auto-encoder embedded molecular graphs into low-dimensional latent space for sampling, a Gaussian Regression Model was employed for prediction and Bayesian Optimization was utilized to select next round of promising candidates• utilized the Graph Convolutional Networks to predict key properties of materials• performed high-performance computing (HPC) on a large dataset of molecules (Python, shell script)

1. Design and Analysis of Storage Tanks in Solar Thermal Power Plants Located in Delingha and Hami∙ conducted the stress assessment, eigenvalue buckling analysis, and fatigue and creep assessment of the molten-salt storage tank under multiple workloads by employing Finite Element Analysis∙ performed the thermal analysis of the preheating and salt-filing processes of the storage tank through Ansys Fluent∙ These storage tanks I designed have been performing exceptionally well, contributing to the Delingha power plant's successful generation of 152.4 million kWh of electricity in 2023.