

Graph Exchange, Fall 2025 - Agentic AI Memory
Graph Exchange is an innovative mini-conference+unconference on Graph-powered AI Agents, Knowledge Graphs, GraphRAG, and all ways where graphs underpin the modern GenAI stack. Brought to you by Neo4j, the category-defining graph database, we’ll review key use cases for graphs and bring together the leading graph experts.
In this Fall Edition, we'll explore AI Memory, that Andrej Karpathy tried to rename as Context Engineering. We'll have several speakers from the Ai Engineer GraphRAG Track, lead by Vasilije Markovic of Cognee, who'll set the stage.
We'll feature a brand new Deeplearning.AI course on agentic graph memory, and you'll come home equipped with the tools to build the most exciting AI startups of tomorrow with agents who understand, remember, and know what they are doing.
We start with the talks setting the theme of Context Engineering and AI Memory.
Vasilije Markovic, Founder, Cognee.
Introducing AI memory and context engineering, with Cognee as an example.
Vasilije is a founder who went back to school to study Cognitive Science to build memory for AI age that works for both agents and humans.
Alexy Khrabrov, AI Community Architect, Neo4j
ASKG: Agent-Server Knowledge Graph
In this talk, we show a knowledge graph of a many MCP servers found online, and how a graph can create a context to reason about servers.Cameron Pfiffer, Developer Relations, Letta
AI, stateful agents, knowledge graphs, Letta, neo4j, MCP, and other such buzzwords
Basically, I'll be showing how to connect a Letta agent to neo4j using MCP. I'll show the agent constructing its own knowledge graph, live.
Cameron is or has been an alpaca rancher, a Stanford PhD, an ultramarathoner, a pianist, and a creator of self-expanding graphs which look eerily like AGI.
Jason Koo, Developer Advocate, Neo4j
An Introduction to the new DeepLearning.AI course on Agentic Graph Construction
More talks coming!
After the talks (a “conference”) part we go to an unconference, splitting unto two circles on reasoning and memory, and discuss use cases and topics that the participants bring and that we vote on to discuss as a group.
"Space is limited and entry will be granted on a first come, first served basis. Arriving early gives you the best chance of joining the event comfortably. Once the space reaches occupancy, no additional entry or re-entry will be possible."