AI Workflow: An end to end AI pipeline

• cs-train: Contains all the data to train the model
• models: Contains all pre-trained saved models for prediction
• notebooks: Contains all the notebooks describing solutions and depicting visualizations
• templates: Simple templates for rendering flask app
• unittest: It has logger test, API test and model test for testing all the functionalities before deploying to production and for maintenance post deployment
• Dockerfile: Contains all the commands a user could call on the command line to assemble the docker image.
• app.py: Flask app for creating a user interface /train and /predict APIs in order to train and predict respectively
• cslib.py: A collection of functions that will transform the data set into features you can use to train a model.
• model.py: A module having functions for training, loading a model and making predictions

Build the Docker image and run it

    ~$ cd AI-workflow
    ~$ docker build -t capstone-project .

Check that the image is there.

    ~$ docker image ls

Run the container

docker run -p 4000:8080 capstone-project

Test the running app

First go to http://0.0.0.0:4000/ to ensure the app is running and accessible.

For training the model: http://0.0.0.0:4000/train

For making predictions using the model: http://0.0.0.0:4000/predict

Reviewing pointers:

Unit tests for the API: unittests/ApiTests.py

Unit tests for the model: unittests/ModelTests.py

Unit tests for the logging: unittests/LoggerTests.py

Run all of the unit tests with a single script: run-tests.py

Read/write unit tests are isolated from production models and logs

APIs for training and prediction: /app.py

Data ingestion automation pipeline: /cslib.py

Multiple models comparison: notebooks/

EDA investigation with visualizations: notebooks/data_ingestion_eda_part1.ipynb

Containerization within a working Docker image: /Dockerfile

Visualization to compare the model to the baseline model: notebooks/time_series_iteration.ipynb