External data & permissions strategies for RAG

Overview:

If you’re a product or engineering leader building AI products, you've probably learned how important external context is for RAG. But ingesting and retrieving your users’ external data comes with many challenges.

We will share the learnings and strategies that our enterprise AI customers use, covering:

How to ingest all of your users’ external data (old and new)
Chunking strategies to optimize for retrieval accuracy
How to model and reconcile third-party data permissions in your app

Recorded on: Dec 4, 2024 | 11AM PT / 2PM ET

Get the recording

Hosted by:

Jack Mu, Paragon Developer Advocate

Sharing our experience as the integration infrastructure for AI companies globally

What was covered

Ingesting users' external data for context

Core concepts we'll cover:

Ensuring you ingest both historical and real-time updates from 3rd-party apps
How to optimize for retrieval accuracy across different data sources
Orchestrating the retrieval process in RAG

Get the recording

Reconciling 3rd-party permissions

Once you've ingested all of your customers' external data, you also need to ensure users can only access the context/knowledge they have access to. We'll cover:

Ingesting & storing permissions metadata from 3rd-party apps (ie. Google Drive)
Determining how often to refresh permissions
How to approach permissions when scaling across multiple integrations and schemas

Get the recording

The embedded integration platform for developers