A meetup talk covering the infrastructure layers beneath AI applications, framed as an 'AI iceberg.' Four companies present their roles: Temporal provides durable execution for AI agents, ensuring workflows survive crashes and resume from checkpoints rather than restarting. Neo4j offers graph-based memory and knowledge storage for agents, enabling semantic similarity search and cached research results. Redpanda handles real-time event streaming at scale for AI workloads, including agent transcript logging and multi-subscriber analytics pipelines. Auth0 (Okta) demonstrates fine-grained OAuth-based permission management for agents, using short-lived scoped tokens to prevent permission accumulation and limit blast radius if an agent is compromised. A shared research agent demo ties all four components together.
Sort: