Capstone Project at Here Technologies

Designed a 3D alignment model integrating Iterative Closest Point (ICP) and feature-based matching, achieving RMSE < 0.01 across 14 map tiles for downstream autonomous driving analytics. Built and tested a Graph Neural Network leveraging geometric and topological features to improve spatial matching and prediction accuracy. Enhanced spatial preprocessing and feature extraction from point cloud data through densification and interpolation, improving geometric consistency and model readiness for large-scale automation.

Data Scientist Intern at Cmind Inc

Engineered automated data preprocessing pipelines in Python and SQL to integrate 10K+ weekly stock datasets, reducing manual effort by 40% and producing high-quality training data. Analyzed multi-dimensional financial datasets using Python (Plotly) and Tableau to uncover short-term performance patterns, anomalies, and correlations, supporting data quality improvement and investment analytics. Standardized data workflows with automated validation and consistency checks, enabling reproducible analytics. Collaborated on NLP preprocessing and data validation workflows across 1,000+ CEO/CFO transcripts, improving data consistency and feature quality. Fine-tuned GPT-based transformer models on earnings call transcripts to extract sentiment and evasiveness embeddings, integrating into a predictive framework for post-earnings stock movement with a 9% performance gain.

Data Scientist Intern, Air Ticketing Department at Ctrip.com

Processed and standardized 1M+ flight timestamp records from multiple data sources using Python-based anomaly detection and time-series preprocessing to improve data integrity and feature consistency. Trained logistic regression and decision tree models to evaluate source accuracy, then developed a dynamic weighting algorithm to combine predictions and improve overall arrival-time forecast accuracy by 8%. Visualized prediction errors and discrepancies across data sources, revealing reliability patterns that informed weighting adjustments and data-quality improvements presented to the analytics team.

Data Analyst Intern at Wangding (Chengdu) Technology Ltd.

Integrated and cleaned multi-hospital healthcare data using SQL and Python preprocessing, generating standardized datasets for modeling and analysis. Applied PCA for dimensionality reduction and risk stratification, visualizing outcomes in Power BI dashboards for 5,000+ patients to assist clinicians in early diagnosis and intervention planning.