According to reports, San Francisco-based startup Niobia is providing production line defect detection software for battery and chip manufacturers. Its AI agents run on standalone GPUs installed directly on manufacturing equipment, capable of identifying defects such as pinhole-sized scratches and aggregating production data for engineers to trace root causes via a chatbot, with the goal of reducing troubleshooting time from days to minutes.

Niobia states that defects can lead to millions in losses and delay production timelines. The company employs AI agents trained further on open-source models to detect subtle flaws in battery and chip manufacturing; given clients' reluctance to upload proprietary data to public clouds, Niobia also enables customers to retain custody of encrypted keys, ensuring the company cannot decrypt customer data. Currently, the company focuses on processes including thin-film coating, electroplating, and fluidized bed reactors, aiming to progressively increase technological autonomy and ultimately achieve fully automated factories.