Hao-Chun Yang

• Conducted research on Unsupervised 3D Representation Learning for robust Anomaly Detection in Medical Imaging. • Improved psychosis detection by 6.7% AUROC through Self-Supervised Masked Mesh Vision Transformer. • Investigated Vision and Language Foundation Model fine-tuning for a large-scale disease detection system.

• Led the development of End-to-End Heart Disease Anomaly Detection system with Time-Series Multi-Lead ECG. • Proposed Mixed-Domain Self-Attention Resnet (MDARsn) to handle missing data and outlier ECG for robust Multi-Label Detection. • Received the Best Challenge Poster award in the PhysioNet/CinC Challenge 2021 and secured a US patent (US20230153575A1).

• (CMedia is one of the top three audio chip suppliers in the world) • Led a team of 5 developing a Real-Time Fully-Convolutional Speech De-reverberation engine, achieving an SRMR score of 4.774. • Oversaw the entire R&D pipeline: Data Collection/Synthesis, Model/Metric Design, Error Analysis, and Model Compression.