Keynote, Adobe Acrobat GenAI Assistant, Robust Security Analytics & Threat Detection
summary
GenAI PDF assistant encodes chunked document JSON into embeddings stored in OpenSearch as a vector DB using KNN with cosine similarity, running real-time indexing and queries with a 12‑hour lifecycle per document via 12-hour rolling indices. Production stats: 11 million embeddings so far, 3,650 embeddings/sec peak, 468 primary shards (936 total) using ~120 GB memory, indexing latency ~395 ms and search ~242 ms. Security XDR pipelines ingest massive telemetry via AWS (API Gateway, Kinesis, Firehose, S3, Lambda) to enable threat detection and correlation across long retention windows.
AI-written summary · about
chapters (7)
- 0:00Welcome to OpenSearch
- 4:44GenAI search with OpenSearch
- 10:47Adobe’s PDF Q&A with OpenSearch
- 16:48OpenSearch for observability
- 19:34Security analytics with OpenSearch
- 23:54QuickHeal threat detection with OpenSearch
- 30:41Closing remarks
related talks
28mUse OpenSearch as Your Vector Search Database To Optimize GenAI Accuracy!
30mSecureAI: A Scalable, Secure, and Compliant AI Solution
41mKeynote
36mGENAI-Powered chatbot to increase team efficiency
22mStructuring the Unstructured : Empowering the AI Workflows
29mCustomizing genAI applications for your business
34mKeynote: Building for AI at Scale: Infra, Integrity, & Innovation
01h 00mComplex Data Ingestion with Open Source AI