• Develop React web applications to manage and analyze foundry data for strategic business planning, improving data classification efficiency by 25%. Implement Azure cloud deployment pipeline with enhanced security measures.

• Analyze historical data and research ML algorithms in Python to predict product issues pre-shipment, reducing failures and enhancing product reliability.

• Designed and developed test framework for Intel Habana Gaudi (AI processor) with Python to seamlessly automate performance tests and smoke tests, saving 2+ days of manual work synchronizing tests during each release.

• Integrated Gaudi performance tests and smoke tests into AWS infrastructure to assist launching new Deep Learning Container on AWS.

• Worked closely with clients to resolve Deep Learning model performance and execution issues on TensorFlow and PyTorch Gaudi backend, which included CNN and transformer models.

• Developed ML framework to enable acceleration of PyTorch and ONNX computer vision models on Cornami AI Chip using C++ and Python.

• Built DL model performance evaluation and analysis tools in Python to increase internal scientists’ productivity.

• Optimized ML operators to tailor Cornami AI chip to accelerate model inference and training time.

• Automated end to end black box testing to ensure the robustness and correctness of the ML framework.

• Researched and developed Neural Network Compression and Representation (NNR) algorithm and integrated the algorithm into modern AI frameworks such as TensorFlow and PyTorch with Python and C++.

• Collaborated with researchers to innovate a compression algorithm capable of reducing model weights up to 20 times for MPEG meetings to compete for the standardization of NNR.

• Developed function to optimize data visualization of monitored servers using JavaScript with Vue and Grafana, as well as built UI components and a complex search function for the web application.

• Tested functions and efficiency of online web application, reported problems, and fixed reported bugs to optimize efficiency of clients’ servers and minimize downtime by 12%.

• Developed Android version of surveillance applications for company clients and published the applications to Google Play, leading to increase customer satisfaction.

• Implemented new video stream analyzing algorithm with OpenCV and FFmpeg to fix backward compatibility issues with legacy cameras, achieving 15% speed improvement in the meantime.

• Built backend services for connecting surveillance cameras and storing recorded videos using Java with Spring and MongoDB, as well as extended push notification services for Android device users in China by replacing default GCM.

• Led a team of 5 as Scrum Master to increase software development productivity and team velocity.