Skip to main content

Enhancing data streaming with real-time data and generative AI

Join this session as we explain why data streaming is a crucial enabler for building responsive, contextually-aware generative AI applications. We discuss why foundation models such as large language models offer immense potential, but lack the ability to dynamically incorporate real-time data at inference time, leading to hallucinations, lack of relevance, and poor personalization. This session dives into the various techniques including in-context learning and retrieval augmented generation (RAG) to help bridge this gap, by allowing models to adapt to the latest data in the context of a given prompt or query. We explore the key architectural patterns you can use when building streaming data pipelines to ingest change data capture (CDC) events, perform identity resolution for unified customer profiles, and transform unstructured content into vectorized representations, all in near real-time. Discover how to use key AWS services including Amazon MSK, Amazon Kinesis, Amazon Managed Streaming for Apache Flink, as well as purpose-built vector databases for building real-time analytics with streaming data.

Speakers:
Partha Sahoo, Senior Analytics Specialist, Technical Account Manager, AWS
Masudur Rahaman Sayem, Senior Streaming Solutions Architect, AWS