Data for AI: One AI, Every Data Model
Last time, we made the case for unifying data so agents can actually use it. This session goes a level deeper, into the structures themselves.
A vector index, an Iceberg table, an object store full of audio and video, a JSON document, a knowledge graph, a stream of events. Each one answers a different question well and fumbles the rest. Force every modality through a single structure, and you pay for it in latency, storage cost, or meaning you can't get back. Choose well, and the same agent gets faster and cheaper.
We'll walk through the various data structures behind multi-modal AI. What each is good at, where each one breaks down, and how to combine them so you don't end up with six disconnected stores nobody can govern.
Built for data engineers, AI/ML engineers, and platform architects building for agents and multi-modal workloads.
Speakers
Bharath Krishna and Akshay Thorat, Senior Engineers at Roku: an AI agent that takes a request in chat and ships an approved merge request on its own. Every hop authenticated, zero hardcoded tokens.
Guy Lubovitch, Director of Customer Engineering at FalkorDB: Store Your AI Agent's Memory and Context in a Knowledge Graph
Mark Hoerth, Product Lead at Datastrato: keeping a multi-model estate governed and queryable through one open catalog, Apache Gravitino.
Vatsal Trivedi, Founder of Runlog Atlas: semantic collapse, what happens when retrieval mistakes noise for relevance, and how the brain's trick of exciting and inhibiting memories inspired their Atlas architecture.
Plus a lightning talk from Kranti Parisa, Founder of LaserData and Apache Iggy (Incubating) PMC member, on Rust-based streaming.
Required for entry
AWS Loft venue policy means every registrant must also sign in through AWS here: http://events.builder.aws.com/d/4dz2b8. Please complete this before the event — you won't be able to get in without it.
All guests must be 18+ and present a valid physical photo ID at check-in (digital IDs will not be accepted). Scooters and bikes are not permitted in any Amazon building with no on-site parking available.
About Data for AI
Data for AI is a community of founders, engineers, executives, and innovators building the data infrastructure behind generative AI, multi-modal models, and whatever comes next. We get together in person to swap hard-won lessons, meet the people solving the same problems, and hang out. Come join us.
Sponsors
This event is made possible by our sponsors, who keep the Data for AI community running.
Datastrato is the company behind Apache Gravitino, the open-source metadata lake that puts tables, models, and files under a single catalog, lineage graph, and access layer. They are building the open data fabric platform to accelerate trusted AI., and host the Data for AI community. More at datastrato.ai.
Neo4j is the graph database behind many of the world's knowledge graphs. Its native property-graph engine and Cypher query language make relationships a first-class citizen, which is why teams reach for it on connected-data problems and, increasingly, GraphRAG for grounding LLMs. More at neo4j.com.
Redis is the in-memory data platform engineers reach for when latency is the whole point. Beyond caching, it now powers vector search, semantic caching, and agent memory, which puts it in the fast path of most AI stacks. More at redis.io.
Want to reach the Data for AI community with your brand? E-mail adi@datastrato.ai