Heidi Lin Email & Phone Number

United States

Berkeley, California, United States

Led a class of 40 students on developing data science projects in industry context, including multimodal recommender system, computer vision and ensemble modeling. (Course: IEOR243 Analyrics Lab)

Bellevue, Washington, United States

• 【Natural Language Processing】
• Developed real-time text classification using deep neural networks (LSTM, CNN) to suggest product categories for sellers.
• Analyzed model failure patterns and edge cases to improve data quality. Implemented pre-trained word embeddings (GloVe) to improve generalization, increasing accuracy by 5% and precision by 10%.【Computer Vision】
• Developed transfer learning for image classification to identify e-commerce images.

Taipei, Taiwan

• 【Machine Learning Development & Productization】
• Built the first in-house credit risk model using XGBoost. Created innovative features from telephone behavior to reduce loan underwriting risk by 50%.
• Deployed the credit risk model to production, automated model retraining and scoring, saving 60+ working hours per month.
• Customized performance and model stability metrics for imbalanced data.
• Designed automated model reports and visualizations using WandB. Implemented Explainable AI using SHAP.【Process Optimization & Analytics】
• Created an EDA tool that combines statistical tests and optimal data binning to shorten data-mining cycle from 7 to 2 days.

• Built underwriting models using logistic regression, increasing loan approval rate for thin credit files by 175%.
• Predicted cross-selling opportunities using Random Forest to unveil and prioritize under-served customer segments.

Taipei, Taiwan

• Devised product features and marketing strategies for a new digital banking service based on customer profiling and profit projection.
• Extracted information from over 1M transaction data, defined metrics and built the data mart from scratch to create a monthly performance dashboard.

Taipei, Taiwan

• Designed and executed A/B testing for SMS marketing campaign, increasing response rate by 160%.
• Proposed new VIP customer segmentation using K-means, projecting customer lifetime value to advise up-selling strategy.

Taipei, Taiwan

• Designed an Excel dashboard with numerical analysis and tendency charts to refine internal utility management process

Master Of Information Management And Systems

Master Of Information Management And Systems (Mims)

• Concentration in Machine Learning and Data Science • Relevant Coursework: Probability for Data Science, Principles and Techniques of.

Bachelor Of Business Administration (Bba), International Business, 3.8/4.0

Activities and Societies: Design Thinking Club; Director of Public Relations; NTU x PKU Elite Camp• Relevant Coursework: Statistics.

Global Bba Exchange Program