Recommendations
Semantic IDs: Product Understanding at Scale
Our short summary of Instacart’s engineering write-up. The full story, with all the detail, is on their blog.
Read the full write-up on tech.instacart.com (opens in a new tab)Problem
Products in different categories never appear together, hurting new and niche items.
Solution
Created semantic IDs by turning product embeddings into short integer codes with a residual vector quantizer, letting similar items share prefixes and be discovered across categories.
Research behind this, on AI Papers DailyUtility-Aware Generative Recommendation: Utility-Enriched Semantic Codes and Utility-Guided Decoding
Want the architecture, code and numbers? They’re in the original post.
Read the full write-up on tech.instacart.com (opens in a new tab)