56454: BYOD Lab: Model Management and Serving on z/OS (Including Model Explainability and Drift Detection)
Project and Program:
z/OS Systems Programming,
New and Innovative Technologies,
Machine Learning/AI
Tags:
Proceedings,
SHARE DC 2025,
2025
Get hands-on experience with the latest AI technologies on the z/OS platform in
this lab, where you will deploy a credit card fraud detection model from
inception to scoring and beyond, walking you through the following stages: 1)
Model training in Python within a Jupyter notebook using the open-source
Tensorflow AI framework; 2) Model exportation in the Open Neural Network
Exchange (ONNX) format; 3) Importation into IBM Machine Learning for z/OS,
leading to the generation of an inference program that taps into the IBM z16's
innovative Integrated Accelerator for AI; 4) Call this z/OS-served inference
program from a z/OS application to obtain a score that determines whether to
accept or reject a purchase; 5) Be able to explain the results for auditors; 6)
detect a drift in accuracy over time. All this will be covered to show how you
can deploy models in this powerful environment.
Back to Proceedings File Library