Scribd
Upload
31 views

Power Amazon Bedrock Applications With Neo4j Knowledge Graph

Uploaded by

dsonam1562
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
31 views

Power Amazon Bedrock Applications With Neo4j Knowledge Graph

Uploaded by

dsonam1562
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 19

Power Amazon Bedrock

applications with Neo4j


knowledge graph
Sumit Shatwara, Sr. Solution Architect, Neo4j India
The State of Generative AI

The Good 💪 The Bad 👎 The Ugly 😱

GenAI Alone != Right Outcomes 🤯


Widespread Hesitancy: Over half of business
leaders currently discourage adoption of genAI.2

Managing AI risk Lack of Explainability: Over 80% of executives


is the biggest worry the non-transparent nature of genAI
could result in poor or unlawful decisions.2
barrier to scaling
AI initiatives1
Risk of Inaccuracy : Inaccuracy and hallucination
are two of the most-cited risks of adopting genAI
technology at all levels of an organization.3

1. Deloitte’s State of AI in the Enterprise 2. BCG’s Digital Acceleration Index Study 2023 3. McKinsey: The state of AI in 2023

4 Neo4j Inc. All rights reserved 2024


The limitations of LLMs

Inability
Hallucinations to verify
answers

Lack of
enterprise
domain
knowledge

Sensitive
Ethical and
to prompt
data bias
(input)
concerns
phrasing
How can organizations use
The Big domain-specific knowledge
to rapidly build accurate,
Question
contextual, and explainable
GenAI applications?

6 Neo4j Inc. All rights reserved 2024


Retrieval-Augmented Generation Is Becoming
an Industry Standard
RAG augments LLMs by retrieving Retrieval Augmented Generation
up-to-date, contextual external data
to inform responses:

● Reduce hallucinations with


verified data
●
Provide domain-specific,
relevant responses
●
Enable traceability back
to sources
Database of Truth

7 Neo4j Inc. All rights reserved 2024


Why RAG With Vector Databases Fall Short
Similarity is insufficient for rich enterprise reasoning

Only leverage a fraction of Miss critical context: Struggle to


your data: Beyond simple
“metadata”, vector databases
alone fail to capture relationships
from structured data
1 2 capture connections across
nuanced facts, making it
challenging to answer multi-step,
domain-specific, questions

3 4
Vector Similarity ≠ Relevance: Lack explainability:
Vector search uses an incomplete The black-box nature of
measure of similarity. Relying on it vectors lacks transparency
solely can result in irrelevant and and explainability
duplicative results

8 Neo4j Inc. All rights reserved 2024


Introducing GraphRAG with Neo4j
Unify vector search, knowledge graph and data science
capabilities to improve RAG quality and effectiveness

Knowledge Graph Data


Vector Search
Graph Science

Find similar documents Identify entities Improve GenAI inferences


and content associated to content and and insights. Discover new
patterns relationships and entities
in connected data
9 Neo4j Inc. All rights reserved 2024
... to Play a Big Role in the Future of GenAI

By 2025, 50% of generative AI


initiatives will have improved reliability
and transparency by combining deep
learning foundation models with
knowledge graphs or other composite
AI elements.”

Source: Gartner®, Technological Implications of Generative AI (August 2023)


GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally, MAGIC
QUADRANT and PEER INSIGHTS are registered trademarks of Gartner, Inc. and/or its affiliates and are used herein with permission. All
rights reserved.
Neo4j & GenAI Reference Architecture on AWS
Data Information
Extraction Knowledge
Sources Knowledge Graph GraphRAG
& Ingestion Apps

Enterprise Knowledge
Search

Structured LLMs Aura Enterprise Edition Decision Support


on AWS System

Amazon Customer Service


Bedrock
Bloom Ticket Triaging
Amazon
SageMaker Recommendations

News Content &


Discovery
Graph DB
Ontologies FAQ Bots

Customer Prioritization
Amazon
Bedrock
Marketing

Assistants
Unstructured Graph Data Science

Step 1: Extract Knowledge Step2: Enrich Step 3: LLM Powered Applications


Demo Architecture

AWS

Input Data

Amazon Amazon
Bedrock SageMaker Application
Studio
Web Interface
User

Neo4j AuraDS Instance

Graph Data Science Graph Database Bloom

12 Neo4j Inc. All rights reserved 2024


Parsing

● Zero-shot with a simple prompt with the

LLM

● Extract SEC EDGAR filing information in

accordance with a Neo4j data model

13 Neo4j Inc. All rights reserved 2024


Chatbot

● Translates English to Cypher

● Consumption using LLM with few

shot prompting

● Data augmentation from Neo4j

response

14 Neo4j Inc. All rights reserved 2024


Semantic Search

If your focus is analyzing documents on a


file system, then vector indexing and search
on text embeddings may be sufficient.

If you need to retrieve and make inferences


about people, places, and things connected
to those documents, knowledge graphs can
help.

15 Neo4j Inc. All rights reserved 2024


Real-World GenAI Breakthroughs
Powered by Neo4j

Leveraging AI Integrating AI with Merging structured Converting vast


for customized knowledge graphs for and unstructured data amounts of unstructured
content at scale smarter supply chain for efficient operations data into actionable
management knowledge

16 Neo4j Inc. All rights reserved 2024


Flexible Cloud Deployment Models on AWS
Graph-as-a-Service Cloud Managed Services Self-Hosted

Fully-managed SaaS White-glove For private


Consumption- managed service by and hybrid
based pricing Neo4j experts cloud, or on-prem

Cloud-native Fully customizable deployment Bring your own license


Self-service deployment model and service levels Full control of your environment
No access to underlying Operate In own data centers Run in any cloud, in your account
infrastructure and systems or Virtual Private Cloud

17 Neo4j Inc. All rights reserved 2024


Getting Started with Neo4j Aura on AWS

GenAI AWS Training


Bundles Partnership Offers

● Pay with Cloud Credits ● Cloud Partner Workshops


Fully-Managed Bundles*: ● Virtual Private Clouds Available ● Hands-on Labs
Starter: (24 vCPU, 128GB) ● Amazon Bedrock ● Graph Academy
Regular: (48 vCPU, 256GB)
Large: (82 vCPU, 384GB)
All bundles include:
● Graph Database
● 70+ AI & ML Algorithms
● 5 user licenses for Bloom for visualization
● Connectors (Data Warehouse, BI, Spark,
etc.)
18 Neo4j Inc. All rights reserved 2024
*Negotiated and Self-Managed bundles also available
Thank you

You might also like