•Worked as algorithm intern at Alibaba Health. Responsible for heart sound (phonocardiogram) classification analysis with machine learning and deep learning methods using 3500 data samples collected from various sources.
•Converted the audio file to spectrograms using Python and built up a Convolutional Neural Network model with Keras and Tensorflow. Applied the transfer learning methods using VGG-16 model. Achieved an overall F1 score of 0.82.
•Deployed the model using Tensorflow Serving Docker to receive and response RPC requests from the client side. Uploaded the service Docker image to Alibaba Cloud’s container cluster.