Tan Yu

Collaborative third-year undergraduate seeking hands-on experience in Computer Vision, Natural Language Processing, or Deep Learning, and eager to expand proficiency in leading ML libraries (TensorFlow, PyTorch, Keras). I am an innovative AI student who enjoys finding practical business solutions through lateral problem-solving approaches. With a firm belief in the revolutionary power of Big Data and Artificial Intelligence, I aim to contribute to groundbreaking solutions for long-standing global challenges. Actively seeking a off-cycle internship in Machine Learning or Artificial Intelligence from 15 January to 31 May 2024.

AL/ML Applications Development Intern@TSMC

• Scouted by Taiwan Semiconductor Manufacturing Company (TSMC) in the Intelligent Manufacture Centre-Computer Integration Manufacturing
• Attained 91% accuracy in a binary classification Computer Vision project by employing a Multi-modal Deep Learning network that effectively combined textual and image data with one-hot encoding and oversampling
• Achieved 83% accuracy in a Siamese network-based multi-category classification Computer Vision project with 16 visually similar categories with data augmentation and transfer learning
• Generated visual explanations using Grad-CAM to understand the decision-making process of a computer vision model for explainable AI, resulting in a 20% improvement in model accuracy
• Detected potential model degradation in 80% of tools sooner than traditional methods by implementing a monitoring system using key performance indicators as well as rolling label, univariate, and multivariate drift
• Utilized machine learning monitoring libraries (DeepChecks, Evidently AI) to analyze data drift's impact on model performance using statistical tests (Earth Mover's Distance, Kolmogorov-Smirnov, Pearson's Chi-Squared, Population Stability Index)
• Applied analytical techniques, such as autocorrelation and spectral analysis (Fourier Transform and Wavelet Transform), to identify and appropriately split periodic data, enhancing data representation and analysis
• Filtered through 500 million raw data points for the appropriate data points using various data cleaning techniques, ensuring data quality and accuracy.
• Calculated key performance metrics, including Process Capability Index and success rate, to evaluate and measure the performance and effectiveness of machine learning models

Bachelor of Science in Data Science and Artificial Intelligence