System Design for Recommendations and Search

Breaking it into offline vs. online environments, and candidate retrieval vs. ranking steps.

System Design for Recommendations and Search eugeneyan Start Here Writing Speaking Prototyping About System Design for Recommendations and Search [ teardown production engineering recsys 🔥 ] · 13 min read Also translated to Korean thanks to Zimin Park. How do the system designs for industrial recommendations and search look like? It’s uncommon to see system design discussed in machine learning papers or blogs; most focus on model design, training data, and/or loss functions. Nonetheless, the handful of papers that discuss implementation details elucidate design patterns and best practices tha

related reading

• How to build production-ready Recommender Systemstheneuralmaze.substack.com
• Building a web search engine from scratch in two months with 3 billion neural embeddingsblog.wilsonl.in
• Tensorflow deep retrieval using Two Towers architecture | Google Cloud Blogcloud.google.com
• The Anatomy of a Search Engineinfolab.stanford.edu
• Personalized recommendations - IV (two tower models for retrieval)linkedin.com
• Patterns for Building LLM-based Systems & Productseugeneyan.com
• Our AI Research: How We Evaluate Semantic Search Technology | Exa Blogexa.ai
• Rerankers and Two-Stage Retrieval | Pineconepinecone.io
• The other hard retrieval problemssoftwaredoug.com
• Beyond “People Also Liked”: Building Intelligent Food Recommenders with LLMs | by Faisal Hussain Sabir | Mediummedium.com
• Introducing Hybrid Search and Rerank to Improve the Retrieval Accuracy of the RAG System - Dify Blogdify.ai
• Recommendation System Series Part 1: An Executive Guide to Building Recommendation System | Towards Data Sciencetowardsdatascience.com