In this talk, I aim to highlight common concerns with building and deploying Machine Learning solutions, and discuss various patterns that can be consistently used to deliver value sustainably and quickly across a broad range of industries and specific business problems.
We will dive into some common patterns such as building composable components for reusability and consistency, production readiness, optimizing for learning with deployed solutions and guidelines for investments in research, as well as discuss tradeoffs against different dimensions such as latency vs complexity, generalizability of modeling solutions vs problem specific optimization. You should walk away from this talk with some general heuristics and patterns that make decisions for your ML solutions simpler, easier to manage and more successful.