Conversational Jobs & Hotels search with metadata-aware embeddings

Описание: In this guest lecture from a Maven course, Daniel Svonava, CEO and founder of Superlinked, challenges conventional approaches to search and retrieval systems. He argues that most businesses are dramatically underutilizing their structured data when building vector embeddings, leading to suboptimal search experiences.

🔍 Key Topics Covered:
Why current RAG approaches focus too heavily on unstructured data
The limitations of text-to-SQL for complex queries with multiple signals
How to encode metadata (timestamps, coordinates, behavioral data) into embeddings
Why re-ranking is "mega overused" and often unnecessary
Building metadata-aware embeddings that understand more than just text
Case studies from jobs marketplace and fashion e-commerce
Evaluation challenges for metadata-rich search systems

💡 Main Argument: Instead of relying heavily on filters, boosting, and re-ranking, businesses should encode as much signal as possible directly into their vector embeddings for more efficient and accurate retrieval.

⚡ Controversial Take: Re-ranking is often just a "hack" to compensate for poor initial retrieval. In an ideal system, the top 10 results from your database should already be the best 10 results.


Timestamps:

00:00 Introduction & Problem Setup
02:00 Travel Query Example Analysis
07:00 Issues with Text-to-SQL Approaches
12:00 Why Stringifying Numbers Fails
16:00 The Case Against Re-ranking
23:00 Custom Encoders vs. Pre-trained Components
28:00 Evaluation Challenges
30:00 Q&A: Agentic Search vs. Metadata-Aware Embeddings
41:00 Manufacturing Use Cases
44:00 3D Geometry Embeddings

#VectorSearch #MachineLearning #RAG #Embeddings #SearchEngines #RetrievalSystems #AI #DataScience