Formulating business problems as AI/ML use cases, finding the right datasets, creating models and fine tuning them. You took your AI Application through a PoC, pilot and customer deployment. You have been there, done that. Great! Now it breaks with new data or in the next geography or next year or at the next customer. What to do? We will share specific techniques through examples we have lived through, to keep the multiple levels of models improving, with novel model generation, automated evaluation, low-cost balanced metrics, and more.
Key Takeaways:
- Identifying what breaks in the field
- Strategies to prevent such breakages
- Techniques for handling challenges such as distribution changes
- Methods to handle lack of direct visibility to business processes
- Explanation of metrics suitable for classification in imbalanced datasets
- Creation of models that handle different regimes of data variability separately
- Sharing our experience in ensuring apps functionality across:
- Different geographical regions
- Varying customer bases
- Business changes
- Different timeframes
- Different geographical regions
In this session you’ll learn how to put it all together for the enterprise. Come share your own experiences and challenges!
