Approach

SageMaker Platform Setup

• Establishment of a SageMaker environment to enable development of a prototype AI/ML model, leveraging SageMaker instances
• Setting up connectivity to data store
• Leveraging SageMaker Model Registry to cataog and manage model versioning
• Leveraging SageMaker Endpoints for model hosting and production
• Enabling additional features such as data labelling, preprocessing, model training, evaluation, and production performance to support a target use case

Prototype Model Build

• Discovery to scope out business problem, objective function, error metrics, potential independent predictors, and a target response for the buildout of an AI/ML model prototype
• Consolidation of data from up to 3 data sources, denormalization into a datastore to enable predictive modeling
• Iterative prototype model built to obtain desired accuracy, including feature augmentation, feature engineering, and model development cycles
• Recommendations for production

Activities

• Setup of the SageMaker environment
• Data consolidation
• Data assessment
• Feature augmentation
• Feature engineering
• Feature importance mapping and assessment
• Iterative model development
• Model selection and hyperparameter optimization
• Results summarization

Deliverables

• Scripts for data consolidation, feature engineering, model development, and assessment
• Scripts for model development and assessment
• Packaged pipelines for prototype model
• Result spreadsheets and visualizations demonstrating model accuracies
• Relevant support visuals, depending on the use case
• Technical workflow summary

Outcomes

• Functional SageMaker environment for machine learning model development
• Understanding of any potential gaps for the final model buildout
• An accurate determination of the investment and anticipated accuracy for a full solution buildout
• Ability to plan for required architecture for implementation