Hack the Video Agent Context Graph (OpenAI + Neo4j + AWS + TwelveLabs)

the vibe

Hack the Video Agent Context Graph Video is one of the hardest data types to search: hours of footage where everything you care about—what is shown, said, and written on screen—is locked inside frames nobody has time to watch. Your challenge is to build a video agent that turns what it sees into a context graph it can reason over. In one full day, your agent will ingest raw video, understand it across vision, audio, and speech with TwelveLabs, and write what it finds into a Neo4j context graph: entities, scenes, tags, and the relationships between them. Strands Agents orchestrates the pipeline. OpenAI drives the reasoning. Go solo or form a team. No prior graph or video AI experience is required. The event operator expects to provide starter templates and the technical stack; final links and access instructions will be shared when confirmed. Prize categories and judging details are still to come. Build something that watches, tags, and connects. What You Will Build With Strands Agents to orchestrate the video-to-graph pipeline OpenAI to reason over extracted context TwelveLabs to understand vision, audio, speech, and on-screen content Neo4j to model entities, scenes, tags, and their connections Schedule 9:30 AM - Doors, check-in, coffee, and setup 10:00 AM - Orchestrating the Pipeline with Strands Agents - Elizabeth Fuentes 10:15 AM - Reasoning over Video Context with OpenAI - Charlie Guo 10:30 AM - Understanding Vision, Audio, and Speech with TwelveLabs - James Le 10:45 AM…