San Francisco-based startup Niobia is focusing on improving defect identification in battery and chip manufacturing. Its AI software is designed to help engineers understand the causes of production issues, aiming to provide answers quickly.
The company's AI agents are deployed on production lines to detect defects, including pinhole-sized flaws. These agents run on self-contained GPUs mounted directly on manufacturing equipment. The AI also centralises data, allowing engineers to use a chatbot to investigate what went wrong and delve into the root causes.
Niobia co-founders Gaurav Jha and Harveen Kaur, both with PhDs in chemistry, have prior experience in chip and battery manufacturing. Jha stated that while the company's work is specialised, it serves a multibillion-dollar market that includes major battery and semiconductor manufacturers.
The AI agents utilise open-source models, which Niobia further trains with battery and chip research data. Jha explained that the choice of open-source models was influenced by customers' reluctance to have sensitive company information used in public AI clouds. Niobia also allows customers to retain their own encryption keys for data security.
Currently, Niobia is prioritising three specialities: film coating, electroplating, and fluidized bed reaction. The company also aims for its technology to become more autonomous, with a long-term goal of enabling fully self-driving factories for its customers.