• Designed, implemented, and deployed an end-to-end pipeline for YouTube RPC response data analysis• Build RPC server for receiving responses, performed data cleaning using self-created C++ library, performed data analysis using parallel data-processing, and logged data into distributed database using C++• Analyzed millions of YouTube RPC responses per day, provided useful information for monitoring feature usage, calculating response data size, and helping to decide on updating or deprecating features of 2+ different teams on YouTube• Increased test coverages and fixed functions for YouTube Development’s interactive testing tool using Python

- Project: EEG (Electroencephalogram) Signal Prediction for Oddball Training Responses- Publication: Speaker-Invariant Adversarial Domain Adaptation for Emotion Recognition, accepted by ICMI 2020 with 38% acceptance rate

- Conducted research on figure classification, used Tensorflow to build and train different CNNand ResNet models- Applied these models into family set-top box for figure recognition- Assisted in detecting distance between human face and digital camera on computer