M.Eng. in Energy and Power Engineering, Tianjin University, Tianjin, China
Sep 2024 – Jun 2027 (expected)
Research focus: AI-enabled materials design for electrochemical energy storage, with a current emphasis on lithium–sulfur battery catalysts
B.Eng. in Materials Science and Engineering, Hebei University of Technology, Tianjin, China
Sep 2020 – Jun 2024
Master’s researcher at Tianjin University working on AI-driven electrochemical materials design, with a current focus on lithium–sulfur battery catalysts. My research integrates machine learning, large language models, density functional theory, and electrochemical domain knowledge. I am the first author of a study on two-stage LLM fine-tuning for lithium–sulfur catalyst discovery published in Materials Today Energy 59 (2026), Article 102371. I have also led or contributed to data-driven research on hard-carbon anodes and single-atom catalysts and have played a leading role in establishing AI-assisted materials research within my group. I am seeking fully funded PhD opportunities beginning in 2027.
My research interest is in AI-driven electrochemical materials design. I have applied this approach to hard‑carbon anodes for sodium‑ion batteries and to single‑atom catalysts for lithium–sulfur batteries. A central motivation of my work is to make models capture the underlying physical laws of materials rather than simply fitting data. Beyond the technical side, I enjoy and excel at leading collaborative research teams; most of my self‑initiated projects have been led by me as the student project lead.
Adapting data processing to model behaviour: I develop Python-based tools to collect, organize, and standardize data from scientific literature. Missing values are initially handled using MICE. During model development, I continually adjust dimensionality reduction, outlier handling, and feature selection according to changes in predictive performance, generalization, and model interpretation. In this way, data processing and model training form an iterative process rather than two separate stages.
Selecting targets that are both scientifically important and learnable: Before training a model, I first determine whether the available data can explain the scientific question. In my machine-learning study of single-atom catalysts for lithium–sulfur batteries, I focused on catalyst stability rather than activity because catalytic activity has already been widely studied, whereas maintaining catalyst function during long-term cycling remains a less resolved practical problem. Activity is also reported through many different measurements and experimental protocols, making it difficult to define a consistent target across publications. By contrast, stability can be described using cycling-derived indicators within a unified evaluation rule. However, the exact numerical stability values are still strongly affected by sulfur loading, electrolyte-to-sulfur ratio, current density, and other cell-level conditions. I therefore converted the continuous target into stability classes, allowing the model to learn broad and reproducible differences between materials rather than fitting small numerical variations caused by testing conditions. The resulting class probabilities or model scores can then be used to rank catalyst candidates and determine which structural directions should be tested next.
Embedding materials knowledge into model development: I integrate materials knowledge into model inputs, training objectives, and output interpretation. In my LLM-based Li–S catalyst study, I defined catalyst-relevant evaluation dimensions and used them to construct the LoRA fine-tuning data. The model produces dimension-specific scores together with supporting rationales, enabling chemically interpretable catalyst assessment. In my machine-learning study, I developed descriptors based on the coordination and electronic characteristics of single-atom catalysts and analyzed their relationships with model predictions. I further compared the SHAP-derived trends with published experimental and theoretical evidence to evaluate their physical and chemical consistency.
Liu, H.; Li, S.; Huang, Y.; Zhao, Z.; Wu, G.; Wang, N.; Sun, Z.; Liu, C.; Yan, H.; Wu, Z.; Wang, Z.; Liu, E.; Shan, Z.; Shi, C.; He, F.; He, C.; Zhao, N.; Tang, Y.; Hu, Y.; Chen, B. (2026). "Data-Driven Discovery of Lithium–Sulfur Catalysts via Two-Stage Large Language Model Fine-Tuning and Mechanistic Validation." Materials Today Energy 59, 102371.
Liu, H.; Li, S.; Zhu, S.; Hu, Y.; Han, X.; Shi, C.; He, F.; He, C.; Chen, B.; Zhao, N. (2025). "How Machine Learning Has Driven the Development of Rechargeable Ion Batteries." Advanced Energy Materials 15, e04095.
Sun, Z.; Liu, H.; Li, W.; Zhang, N.; Zhu, S.; Chen, B.; He, F.; Zhao, N.; He, C. (2025). "Advanced hard carbon materials for practical applications of sodium-ion batteries developed by combined experimental, computational, and data analysis approaches." Progress in Materials Science 149, 101401.
Ongoing research, 2026
Oral Presentation at Beijing-Tianjin-Hebei Clean Energy Conference, Tianjin, China