◎ Responsible for developing Text-to-Speech systems(TTS) and low-latency speaker conversion systems using the PyTorch deep learning toolkit.◎ Develop dynamic-link library and console application using C/C++.◎ Integrate libtorch and OpenVINO to deploy machine learning model inference on personal devices.◎ Development of Text-to-Speech system for ALS people. - Reduce the text-to-speech system’s data requirement from 3 hours to 30 minutes.◎ Automated text linguistic parameters and prosodic annotation system.▷ Taipei City Government Entrepreneurship Grant - Reduce the pre-processing time of the data required for deep learning and greatly reduce the errors that can be caused by manual work.◎ Fansi – Unrevealed mastered tape NFT of Tom Chang(張雨生). - Train a deep learning model using the 30 minutes interview transcript preserved by Tom Zhang and synthesize the requested speech content for Fansi.

◎ Responsible for developing Text-to-Speech System for aphasia people. - Corpus design and collect corpus from aphasia people. - Annotate linguistic and acoustic features from corpus for deep-learning. - Train and deploy deep learning models.◎Participate in development of Speech Labeling and Modeling Toolkit(SLMTK) - Responsible for developing deep-learning automated training and deployment modules using Python and PyTorch. Also participating in feature extraction and signal processing of corpus using C/C++.