● Led the migration of a React-based website to Next.js, optimizing it for server-side rendering (SSR) capabilities and integrating API routes for enhanced performance and security.
● Responsible for implementing and managing Stripe Checkout for international subscription, deposit, and service payments.
● Designed and managed service scheduling workflow.
● Designed user document upload and management through AWS and Supabase.
● Researched and built Pytorch image classification models for IAD and PI to facilitate digital tools transition within the trauma reconstruction department
● Trained EfficientNet model on a limited and unbalanced dataset, received a 0.90 recall rate
● Built back-end service with Django and SQLite to provide a user interface for web browser and Line bot, increasing the overall workflow efficiency for data collection and analysis by 80%
● Conducted research and developed PyTorch models for speaker verification systems, text-to-speech technology, and continual learning
● Improved the original model for better stability and accuracy by changing the architecture of the network, the loss function, and the transformation function of the spectrogram
● Used Python and Pandas to improve data quality and constructed a patient data matching software which helped organize over 20M patients and their updated information from over 50 hospitals in Taiwan
● Reduced overall working time for information integration from months to days