San Francisco-based startup Niobia has developed AI software designed to detect production defects in battery and chip manufacturing. The system utilizes AI agents installed directly on manufacturing equipment via self-contained GPUs to identify flaws, such as microscopic scratches, in real time.
The platform integrates production data, allowing engineers to consult a chatbot to determine the root causes of failures within minutes. This process aims to reduce the downtime typically required for manual troubleshooting, which can otherwise last for days and result in losses reaching millions of dollars.
Niobia utilizes open-source models trained on specialized research data to ensure data privacy. To address security concerns regarding proprietary information, the company allows clients to manage their own encryption keys, preventing Niobia from accessing or decrypting sensitive manufacturing data.
The startup currently focuses on three specific manufacturing processes: film coating, electroplating, and fluidized bed reaction. Looking ahead, the company intends to increase the autonomy of its technology with the long-term goal of enabling fully self-driving factories for its customers.
