Topic: AI4Science – Low-Rank Embedding Physical Network: Data assimilation of uncertainty using LoRA in surface low temperature prediction tasks. Designed and implemented a digital twin framework for the roadbed and pavement thermal conduction model, allowing for easy definition of the number of layers, interlayer thickness, and material properties of the roadbed and pavement. Discretized the Fourier equation with the linear finite difference method, constructed the physical numerical model using the PyTorch code framework - parallel program based on the AI Library was 4 times faster than the CPU program. Embedded LoRA network in the physical model to capture uncertainty in the physical system through data assimilation.

 Devised an ItemCF collaborative filtering model for product recommendations with manager ratings, prognosticated client product preferences and advised about the top 10 unowned products per client. Evaluated dormant corporate client data to formulate a model with 77 features; refined 7 key variables, applied MinMax scaling, and utilized optimized K-means clustering to identify 8,221 high-value clients for activation. Innovated, under guidance, a UiPath robot to automate customer data extraction and querying on the Industrial Bankplatform, bolstering efficiency with batch-saving mechanisms and amassing key customer metrics from vast datasets.

Project: Automobile Brand Pricing Regression Analysis Utilized SPSS and Python to conduct regression analysis on 100 automobile brand pricings, identified 3 potential influential factors, of which 2 had significant implications on pricing. Used Tableau to explore 5 major car brands, synthesized the strength and weakness of each brand's pricing strategy.Project: Research on the Vehicle Maintenance Platform Brand Tuhu Used Python to research China's automotive aftermarket, identified an industry growth rate of 4%. Conducted a comparative analysis of business models between Chinese and US markets using Tableau, visualized main differences and assessed their potential implications on the automotive sector. Performed a time series analysis of Tuhu’s development from 2019-2021 using Python, evaluating 20 data points, pinpointing target consumer groups and the brand’s performance in major Chinese cities.