Led machine learning team to develop prediction and optimization solutions based on time series data.

• Defined project scopes, architecture, and functional specification of machine learning solutions.

• Architected machine learning pipeline and led 4 machine learning engineers to deliver a form recognition solution by integrating ResNet, OCR engines(Tesseract/Google Vision), and NLP model for barcode detection, information extraction on the AWS EC2 platform as a Restful service with 89% accuracy.

• Led data engineering team to define data standards, and build the data ingestion, conversion, validation, and preprocessing pipeline for real-time IoT streaming data in AWS S3 solutions by Python.

• Enhanced industrial solution for advanced process control pipeline with data preprocessing (outlier removal, smoothing...etc.), deep learning model (sequence to sequence) as an estimator, and the genetic algorithm as a multi-objective optimizer to generate operation prescriptions based on constraints, objectives for the asset.

• Enabled researches on Capsule Nets and SegCaps image segmentation tasks in Tensorflow.

• Developed a crawler program and data pipeline to collect data from the COCO dataset. Coded a package for 3D computed tomography scans and 2D color images on binary semantic segmentation in Python, Keras.

• Built a video segmentation program with a pre-trained SegCaps model in Python.

• Built Recurrent Neural Network (RNN) to predict the LA freeway speed with Root Mean Square Error (RMSE) 3.1 miles per hour from over one million time-series data points by regression LSTM in Keras.