Use Case 1:
π PDF Search App with Vector Search and LLMs
πΌ Retrieval-Augmented Generation (RAG)
Problem: π
- PDF is unstructured data
- Difficult to search
- Leading claim adjusters reviewing contracts, claims
- Need efficiency and accuracy for this historically cumbersome task
- Summarize content of documents
- Indicate source of information
Challenge: π€―
- Underwriters, claim adjusters go through numerous pages of guidelines, contracts, reports
- Time-consuming
- Lead to expensive mistakes (incorrect risk estimations)
Details: ποΈ
- Provide PDF based on user's country
- Vector search query particular document
- Filtering out the non-relevant PDF
Index Pre-Filtering: π
- Must index fields with data type
- Only includes record with correct data type
Search Nodes Dedicated Architecture: π οΈ
- Support both Atlas Search & Vector Search workloads
Use Case 2:
π Claim Management using LLMs and Vector Search for RAG
Challenges: πββοΈ
- Claim adjusters pulling and aggregating info from different systems with data formats
- Content-sharing platform
- Customer information locked in a legacy CRM
- Claim-related pictures, voice reports
- Unstructured data
- Streamline operations
Solution: ππ€
- Atlas Vector Search
- Large Language Model (LLM)
- Retrieval Augmented Generation (RAG)
Vector Search Query: π
- Get both vector embeddings and plaintext metadata
- Don't need to retrieve data
Aggregation Pipeline: π§
- Process multiple documents and return computed results
Reference:
Claim management using LLMs and vector search for RAG
https://www.mongodb.com/solutions/solutions-library/claim-management-llms-vector-search
Harnessing the Power of Atlas Vector Search and Open Source Models
https://www.mongodb.com/developer/products/atlas/building-generative-ai-applications-vector-search-open-source-models/?hideMenu=1
Build a PDF search application with vector search and LLMs
https://www.mongodb.com/solutions/solutions-library/pdf-search-with-vector-search-and-llm
Editor
Danny Chan, specialty of FSI and Serverless
Kenny Chan, specialty of FSI and Machine Learning
Top comments (0)