The Aryn Journey - Growing a Gen AI Startup on OpenSearch and the Road Ahead
summary
Aryn builds a GenAI stack on OpenSearch, adding a RAG search processor and conversational memory to enable retrieval-augmented conversational search; the project ships in 2023 and OpenSearch hits ~1M downloads/week by early 2022. Aryn runs a cloud parser for ~30 document types plus an open source transformer called Sycamore to extract metadata and load highly accurate vectors into OpenSearch and Amazon OpenSearch Service. The bet: optimize OpenSearch for agent throughput (aiming at 60,000 QPS) as token inference speeds jump from ~20 to hundreds–thousands TPS.
AI-written summary · about
chapters (9)
- 0:00Aryn journey intro
- 1:06Why OpenSearch for startup
- 3:38Generative AI and unstructured data
- 4:43RAG features in OpenSearch
- 6:31Enterprise customer needs
- 8:14Deep analytics agents on OpenSearch
- 10:01OpenSearch AI adoption growth
- 11:27OpenSearch future for agents
- 18:34Closing and Q&A
related talks
25mKeynote: OpenSearch in the Agentic Era
43mHarnessing DeepSeek-R1 and OpenSearch for RAG-Powered Intelligent Info...
30mOpenSearch for GenAI Storage: AgentScope Studio Practice
26mOpenSearch: Conversations with Context
06mSponsored Keynote: OpenSearch - See Everything: Open Observability for Agentic AI
30mNext Gen Agentic Search on AWS
30mBeyond Keywords: Building an AI-Powered Search That Reads Your Mind
47mBoosting AI With OpenSearch: Vector Stores, Similarity Search, and Retrieval-Augme...