Machine learning (ML) has gained tremendous traction in research and industry over the past few years. Its remarkable progress, supported by achievements in Big Data architectures, has enabled the creation of practical cutting-edge AI applications. Once available only to scientists, today ML is achievable by a much broader audience of software architects and engineers. In fact, the practice of ML is so advanced that some algorithm results look like “black magic,” even for experienced practitioners.
But how do we design such systems? Best practices in software architecture guide us with robust methods to analyze requirements and create architecture designs that predictably satisfy business and system needs. Inspired by attribute-driven design (ADD) and Smart Decisions (a software architecture design game for Big Data), the presenters are happy to introduce a new version of the game focused on designing ML systems.
In this participatory session, you will have fun and learn about designing the architecture for ML systems via a series of gamified interactive exercises. We will simulate the state of the art of designing ML systems through analyzing business and technical requirements, selecting the best matching algorithms, and validating early design decisions using rapid prototyping techniques.