Niobia Uses Localized AI Agents to Diagnose Chip and Battery Defects

San Francisco startup Niobia has developed an artificial intelligence platform designed to help manufacturers identify and diagnose production defects in batteries and semiconductors, TechCrunch Startups reported.

The company's architecture places AI agents directly onto manufacturing lines using self-contained GPUs mounted on equipment like sensors. These localized units visually scan for pinhole-sized defects, such as scratches, while aggregating operational data into a central system. When a flaw is detected, engineers can query an AI chatbot to determine the root cause, aiming to cut diagnostic time from days down to minutes.

Unplanned defects during research and production can cause delays and cost millions of dollars in wasted materials, particularly in supply-constrained chip and battery sectors. Niobia was co-founded by Gaurav Jha and Harveen Kaur, who both hold doctorates in chemistry and previously worked at semiconductor and battery manufacturing companies. The startup presented its technology as a finalist in the Startup Battlefield competition at TechCrunch Disrupt 2026.

To address industrial security and proprietary data concerns, Niobia trains open-source AI models on specialized chemical and hardware research and runs them locally rather than relying on public cloud infrastructure. Customers also maintain their own encryption keys so the startup cannot decrypt their data.

Niobia is initially targeting three manufacturing processes: film coating, electroplating, and fluidized bed reactions. Over time, the company aims to expand the system's autonomy toward fully automated factory operations.