3 key steps to effectively scale AA and AI for technology transformation
Interactive
At ENBD, what started as a centralized data governance and management approach to help gather key data assets for early use cases then shifted to a data mesh approach.
1
2
3
Decentralize data gathering
Automate model development and deployment
Keep models relevant and accurate
Scaling data and technology effectively across an organization requires a strategic big-picture plan with three key components.
This meant enrolling data producers across the organization to curate data. The result is a decentralized approach to data gathering that allows relationship managers to easily manipulate data for specific business use cases.
1
Decentralize data gathering
2
Automate model development and deployment
To operate at scale, model development and deployment needs to be automated. At ENBD, this meant shifting from the constraints of an existing software development lifecycle to a machine learning development lifecycle, which delivered models more quickly and enabled safe scaling through automation.
Containerization, a process that virtualizes operating systems so that applications can run more easily in isolated user spaces, was also used. Machine learning operations were put in place to help streamline the process of producing, maintaining, and monitoring different models. All of this streamlined the time and effort it took to engineer new features and helped create a catalogue of reusable assets.
3
Keep models relevant and accurate
To ensure models stayed relevant and accurate, model validation and monitoring were also automated. A model validation framework was put in place to automate key functions like standardizing model documentation, checking for simple logics, registering models, and capturing feature lineage.
A model management framework was set up to observe the pipeline of data and machine learning in real time.
Reset
Step 1: Decentralize data gathering
Next step
Step 2: Automate model development and deployment
Next step
Step 3: Keep models relevant and accurate
Get started
Decentralize data gathering
Automate model development and deployment
Keep models relevant and accurate
1.
2.
3.