Workflow · Production
What Is Machine Learning (ML)? — Dataiku educational guide
The problem
Organizations struggle to deploy ML models to production due to IT productionalization gaps, lack of business-aligned objectives, and data preparation consuming up to 80% of project time.
Workflow diagram · grounded in source
1
Data Access and Preparation
trigger
“This begins with data access and preparation (including feature engineering), and then involves the design, training, exploration, and selection of the best model”
2
Model Design and Training
ai_action
“An ML model is trained to identify the specific patterns using a reference set of data. This training results in an ML model that can be used on future datasets to identify similar patterns and make predictions”
3
Deployment to Production
integration
“This process typically involves readiness steps to first stress-test the model, analyze it for fairness and potential bias, and gain the appropriate approvals. During this phase, the model is deployed to a QA or production environment an…”
4
Performance Monitoring
feedback_loop
“set up the feedback loop required to monitor the model over time to detect drift or declining performance in production”
5
Model Retraining
feedback_loop
“Predictive models will need retraining when the data or conditions in the real world differ from what the model was trained on. This data drift happens over time and can impact predictions' accuracy, resulting in the need to retrain the …”
Reported outcome
With Dataiku, IT teams no longer need months to recode workflows, and all data work can happen in one tool with version control, transparency, and collaboration.
Reported metrics
Data preparation share of project timeup to 80%
Reported stack
DataikuAutoMLScikit-learnXGBoostMLLibSparkSalesforce
Frequently asked questions
What did this team achieve with this AI workflow?
With Dataiku, IT teams no longer need months to recode workflows, and all data work can happen in one tool with version control, transparency, and collaboration.
What tools did this team use?
Dataiku, AutoML, Scikit-learn, XGBoost, MLLib, Spark, Salesforce.
What results were reported?
Data preparation share of project time: up to 80% (source-reported, not independently verified).
How is this workflow AI workflow structured?
Data Access and Preparation → Model Design and Training → Deployment to Production → Performance Monitoring → Model Retraining.