Yuqi Lei

Master's

Computer Science

OPPO Mobile Technology

algorithm engineer

• Android Development: Developed an Android APP using Java & C++ to deploy the neural network model on mobile terminals. Designed the user interface to not only select and test the model, but also to show the time and resource consumption of each model's execution phase, helping engineers make better optimization strategies. Realized C++ APIs to initialize, process, and release the TensorFlowLite model, and use Java Native Interface(JNI) to call these methods to facilitate subsequent model porting. • Edge Computing Model Deployment: Contributed to deploying an edge computing project on mobile terminals with the USTC using C++. Designed a class to manage the communication process between the terminal and the cloud to improve efficiency by reducing unnecessary memory duplication. • Inference Speed Promotion: Improve the inference efficiency on the OPPO self-developed chip by reconstructing C code related to model inference. Fixed byte alignment bugs in some operators to reduce their space consumption by 30%.