How to Build a Full-Stack Recommender System

Jacopo Tagliabue, RecSys expert and former Director of AI at Coveo, will join Hugo Bowne-Anderson, Outerbounds’ Head of DevRel, in our first live code along session to dive into how to build a production-grade RecSys, the goal being to develop a relatively simple, effective, and general pipeline for sequential recommendations. We’ll show how you can use popular open-source libraries and tools including DuckDB, Gensim, Metaflow, and Keras to build a fully working cloud endpoint that serves predictions in real time, starting from raw data. We’ll be using the Metaflow sandboxes so you can easily code along (think Colab but for MLOps)! Using the Spotify Playlists dataset, you’ll learn how to - take a recommender system idea from prototype to real-time production; - leverage Metaflow to train different versions of the same model and pick the best one; - use Metaflow cards to save important details about model performance; - package a representation of your data in a Keras object that you can deploy directly from the flow to a cloud endpoint with AWS Sagemaker. 00:00 Prelude 04:55 How to code along: Metaflow sandbox (Colab for MLOps)! 12:56 Intro to RecSys 27:43 Building a neural network sequential recommender system live 28:50 Understanding the data: Spotify playlists 37:48 Query faster with DuckDB 47:40 Build a recommender model 51:20 Tuning hyperparameters for model performance 59:30 Analyze results and iterate 1:03:11 Deploy an endpoint to make real time predictions