Verkada’s AI-powered search leverages a hybrid cloud architecture, combining edge-based camera processing with scalable AWS backend compute to enable fast, flexible investigations. Using the OpenCLIP model for zero-shot image-text matching, Verkada transforms user queries into relevant results from hyperzoomed (HZ) image segments. In-house enhancements like output caching, query parsing, and dynamic thresholding improve speed and accuracy. A robust moderation system filters inappropriate content to ensure safe usage. Though still evolving, this AI search system delivers high recall and scalable performance while maintaining privacy, user control, and ethical safeguards.

Join for free to read