Overview / Usage

tensorflow is a popukar framework for developing ML applications and there is indeed a need for an understanding in the marketplace. Because there are appropriate tutorials and demo codes available at tensorflow.org, the initial step calls for familiarizing with those (i.e. learning by doing):

• single-node/mutli-node jobs
• different algorithms including deep networks, GANs, regression tools, etc
• Python binding
• production-acceleration features such as estimators, keras, etc

Methodology / Approach

• use codes available at
  https://www.tensorflow.org/guide/custom_estimators
  https://www.tensorflow.org/guide/premade_estimators
  https://www.tensorflow.org/guide/datasets_for_estimators --> explains the input function
  and other github repositories in order to better learn tensorflow/python features
• develop my own tensorflow demo package for potential customers
• refer to python documentation as e.g. https://upload.wikimedia.org/wikipedia/commons/9/91/Python_Programming.pdf

Technologies Used

tensorflow, tensorboard, keras, python
later experiment with horovod https://eng.uber.com/horovod/