Analytics Vidhya
Published in

Analytics Vidhya

Why Data Science development is different from Software or Data/BI development


  • Bring production or actual data from various sources
  • Data engineer the data set for use case
  • Build a working data set for modelling
  • Run data set through various algorithms
  • Need large compute depending on use case
  • Compare the performance of model
  • Iterate through the process to find the best model outcomes
  • Once the algorithm is found, Create Training script
  • Create a model file
  • Validate the model with new data set to test the performance
  • if the performance is acceptable move forward, other wise go back to model development
  • Once acceptable performance is good create score script for realtime or batch inference
  • Create Evaluate or model compare script from previous run
  • Scoring script to check for model performance to trigger new model build
  • Save all the scripts in Github or some code repository
  • Create documentation on use case, model and it’s usage

QA and Production

  • Get the code from code repository
  • Create Azure DevOps or other pipeline tools to build the deployment process
  • Run the Training process with production data
  • Need large compute depending on use case
  • Run testing of the model
  • if the model performance is better than previous run then create Model brain file
  • Use the Score file to create REST Api
  • Create Docker container to run the Microservice
  • Create the container orchestration enginer (if new), other wise use existing
  • Deploy the new Rest API and decommission the old one (if exists)
  • When score model is used if performance degrade create a model train run and deployment

Originally published at



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store