1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.A I M 2 1 4-SRAG is Dead:Long Live Intelligent Retrieval-Augmented GenerationArjun Patel(He/Him)Senior Developer AdvocatePinecone 2025,Amazon Web Services,Inc.or its
2、 affiliates.All rights reserved.Serving a good recommendationWhat is Agentic RAG?Query UnderstandingRetrieval ImprovementsGenerating BetterLearn more with Pinecone 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Imagine youre building a tech startupAnd you want to help new couples
3、 find first-date restaurantsGiven user query,you find dishes and restaurants-Avoids dairy-Only eats chicken-Prefers spicy dishes-Loves fish and beef-Values a romantic atmosphere-Specific food allergiesPerson 1 PreferencesPerson 2 Preferences 2025,Amazon Web Services,Inc.or its affiliates.All rights
4、reserved.“We need date night restaurants within our budget near us that has vegetarian nondairy options,spicy options,maybe a good shareable dessert with a romantic vibe”2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Imagine youre a food-tech startup.And this is a real problem pe
5、ople face!You only get one shot.Its me!My fiance,Divya 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Why is this problem so hard?Restaurant reviews,descriptions,dishes,images,geolocation,dietary restrictions.Natural LanguageQueryThe query is long and requires reasoning,planning,
6、and coordination to solveHigh Risk RewardApp-user acqusitionRestaurants-2 seatsCouple-successful dateRAG wont workQuery is just too complex to throw out our databases.We need more engineering workDistinct Data Sources 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web