Using NVIDIA AI Blueprint for Video Search and Summarization with Reka Vision Agent

We are proud to announce the integration of our Reka Vision Agent with NVIDIA AI Blueprint for Video Search and Summarization (VSS). This milestone deepens our longstanding partnership with NVIDIA, where we have used Megatron-Core to train our multimodal models from scratch.

The new integration allows companies who adopt the video data processing pipeline of NVIDIA AI Blueprint for VSS to use our proprietary AI models for accurate natural language search, question answering, and event detection.

Reka Vision Agent is our agentic AI software that understands the physical world. It excels at activity understanding and object tracking across multiple real-time or prerecorded videos — delivering functionality that outperforms visual language models such as GPT 4o. 

Real-world examples include: 

• Summarizing footage with timestamps of “a person walking with a white dog opening a backpack” or “people fighting in the cafeteria“. 

• Generating a comprehensive report of “safety violations that happen in our factory within the last 24 hours”.

These capabilities unlock a smarter, faster, and more intuitive video understanding at enterprise scale. 

A major challenge for enterprises today is to integrate next-generation AI models (with flexible natural language search and cross-camera tracking) without overhauling existing systems. NVIDIA AI Blueprint VSS standardizes low-level inputs and high-level outputs of video analytics systems. This means that we can integrate our proprietary AI models for easy large-scale deployments.

With Reka Vision Agent and NVIDIA AI Blueprint for VSS, customers do not have to worry about compatibility with their existing solutions. Contact us today for a personalized demo of Reka Vision Agent.