Personalized Recommendations at Scale
Surfacing relevant content from among millions of candidates to users in real-time is a challenging task addressed by recommender systems. Most modern-day recommender systems rely on the complex interplay between different components, each of which is powered by sophisticated machine learning algorithms. In this course, we provide a holistic overview of ML modeling choices that go into developing and deploying multi-stage recommenders capable of serving recommendations from hundreds of million content choices to multiple hundred million users. The course goes into algorithmic models that power the various stages of the recommender, including the candidate generator, core ranker, user representation learning modules and offline and online evaluation module. The course ends with case studies, lessons and practical considerations from deployed systems powering over 400 million users.
Course taught by expert instructors
Director of Machine Learning at ShareChat
Rishabh Mehrotra currently works as a Director of Machine Learning at ShareChat based in London. His current research focuses on machine learning for marketplaces, multi-objective modeling of recommenders, and the creator ecosystem. Prior to ShareChat, he was an Area Tech Lead and Staff Scientist/Engineer at Spotify where he led multiple ML projects from basic research to production across 400+ million users. Rishabh has a PhD in Machine Learning from UCL, and 50+ research papers and patents. Some of his recent work has been published at conferences including KDD, WWW, SIGIR, RecSys, and WSDM. He has co-taught a number of tutorials and summer school courses on the topics of learning from user interactions, marketplaces, and personalization.
Learn and apply skills with real-world projects.
- ProjectImplement 1-3 candidate generators, from simple recall-based CGs to noise contrastive estimator-based CGs. Conduct recall-based evaluation of the different approaches.Learn
- Aspects of recommendations & personalization
- Deep candidate generation techniques: negative sampling, contrastive learning
- System design example of few large scale recommenders
- ProjectImplement a multi-task recommender, predicting multiple user engagement signals, and combining various predictions to serve the top-k recommendations. Perform various ranking and user engagement-based evaluations.Learn
- Classical approaches of recommendations: embedding, matrix factorization, factorization machines
- Multi-task recommenders: predicting multiple facets of user engagement
- [Optional deep dive] Sequential recommenders: sequence aware approaches, in-session personalization
- ProjectImplement two user representation techniques and compare their performance on downstream recommendation tasks.Learn
- Learning user representations: leveraging user interactions to learn user representations
- Learning user intents: methods to understand & predict user intent
- Learning from feedback: online methods of feedback incorporation
- ProjectUsing logged data of user interactions, implement a few session-based metrics, and identify the biases of the metric.Learn
- Beyond AUC: offline evaluation setup, session level metrics, implicit signals logged data
- Nuances of online experimentation: AB test design, and online metrics
- Offline-online correlation: scaling
- Counterfactual evaluation: randomized data collection, unbiased estimators of metrics
Work on projects that bring your learning to life.
Made to be directly applicable in your work.
Live access to experts
Sessions and Q&As with our expert instructors, along with real-world projects.
Network & community
Core reviews a study groups. Share experiences and learn alongside a global network of professionals.
Support & accountability
We have a system in place to make sure you complete the course, and to help nudge you along the way.
Course success stories
Learn together and share experiences with other industry professionals
This course is for...
Industry practitioners tasked with developing and deploying a large scale recommender system
MLEs with prior machine learning experience looking into diving deep into large scale recommender systems
Some familiarity with basic machine learning concepts like model training, feature representations, labels
Ability to write Python and work with documented libraries