This conference talk explores how Uber Eats overcame significant scaling challenges when indexing millions of stores and billions of items while maintaining seamless user experiences across discovery surfaces. Dive into architectural strategies including latitude and hex sharding techniques, custom index layouts optimized for different search patterns between Eats and Grocery services, and the implementation of Estimated Time of Delivery (ETD) indexing with parallel query execution to solve critical latency issues. The 49-minute presentation provides valuable insights for engineers facing similar scaling problems in search and discovery features, with a complete transcript available on the InfoQ website.