Case Study: NATIX achieves scalable edge-case mining for Physical AI with NomadicML

NATIX x NomadicML - Scaling Edge Case Mining for Physical AI

NATIX operates a decentralized camera network that collects large-scale, multi-camera driving video. Their challenge was that the rare, safety-critical events in their vast dataset were too sparse and complex to find using signal-based triggers or manual review, making long-tail discovery for Physical AI extremely difficult. NomadicML provided its video intelligence service to transform this raw video into a searchable system.

The solution from NomadicML involved a three-step pipeline using semantic search and vision-language models to retrieve and validate specific edge-case scenarios. This allowed NATIX to query their archive for complex events like near-misses and lane encroachments, reducing search time from hours to seconds. The results included scalable, high-recall discovery of rare events and the creation of structured, model-ready data, forming a continuous learning loop that improved both NATIX's intelligence engine and NomadicML's own models.

NATIX

Alireza Ghods

Chief Executive Officer