Building video context systems for the agentic era.
Overcast: Video OSINT Agent. Point It at 100 Videos, Ask Anything
A CLI agent and skill pack that drops into any agentic harness and turns footage into cited evidence — speech, on-screen text, objects, faces, and named entities, all accumulating in persistent case memory.
🗓️ Fri Aug 7, 12:00 & Sat Aug 8, 16:00 PDT · LVCC Exhibit Hall West 3 (Demo Labs Track 6) · github.com/kdr/overcast
Evidence-Grounded Video Investigation over Large Video Corpora
🗓️ Demo Session TBD · Sept 10–12 · Malmö, Sweden
- Autonomous Video Hunter — AI agents for real-time OSINT · DEF CON 33 Recon Village · watch
- AI That Gets the Picture — building video-aware assistants · CascadiaJS 2025 · watch
- MLOps at Snapchat — continuous ML with Kubeflow and Spinnaker · KubeCon + CloudNativeCon · watch
- SageMaker Ground Truth — building accurate training datasets · AWS re:Invent 2019 · watch
Published at NeurIPS · AAAI · ICCV · CVPR · WACV · KDD · WWW · CAIS · SIGIR · RecSys · NAACL · ISMIR