Are You Ready to Engineer and Sustain AI Systems?
Abstract
Contents include: AI at CMU and AI at the SEI; AI-enabled systems are software systems!; Can we Design and Analyze AI-Enabled Systems Predictably?; Architecture Challenge #1: Lack of Systems Perspective; Recommendation: Manage AI Component, Data, and Architectural Dependencies; Architecture Challenge #2 : Inability to separate data and system attributes; Recommendation: Understand High-Priority Quality Attributes of ML-Enabled Systems; Architecture Allows Improving Predictability of Data and Other System Component Interactions; Architecture Challenge #3 : Lack of Monitorability; Recommendation: Decouple Different Aspects of Monitorability; Recommendation: Integrate the analyses performed by the Data Scientist into the MLOps pipeline; Architecture Challenge #4 : Different Paces of Change; Recommendation: Embrace Changing Anything Changes Everything Principle*.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 11, 2022
- Accession Number
- AD1185241
Entities
People
- Ipek Ozkaya
Organizations
- Carnegie Mellon University