Jeremy Adams-Casañas

Adams’ speaker biography lists work at Neo4j, Dagger, GitHub, Twistlock and Puppet. It describes a mix of technical and strategic responsibilities, with a recurring interest in integration and automation.[2]

For a video-agent hackathon, Adams adapted William Lyon’s create-context-graph starter into a video-focused application. The stack combined TwelveLabs for video understanding, Neo4j for the context graph, AWS Strands for orchestration and OpenAI for reasoning. His write-up explains how teams used that common starting point for recipe extraction, rehearsal feedback and animal-behavior analysis. The starter reduced setup work so participants could focus on their application’s graph and behavior.[1]

The video starter stores each video as a sequence of segments, with links to entities and topics. Entities with the same normalized name merge across videos, allowing related evidence to accumulate around one node. Segment embeddings support similarity search, while the interface displays the graph and a video inspector beside chat. That structure lets a developer trace an answer back toward the material the agent used.[4]