# Niobia Deploys On-Device AI to Find Battery and Chip Defects

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/31444
Published: 2026-10-13T19:06:21.000Z
Updated: 2026-10-13T19:06:21.000Z
Section: Technology

> The San Francisco startup uses AI agents running on open-source models trained further on battery, chip and research data to help engineers investigate production flaws.

San Francisco startup Niobia is developing AI software that helps battery and semiconductor manufacturers detect production defects and investigate why they occurred, targeting a specialized, multibillion-dollar industrial market.

The company places AI agents on manufacturing equipment, where they can identify flaws as small as pinhole-sized scratches. Self-contained graphics processing units mounted directly on the equipment run the systems, while Niobia gathers the resulting data in one place for engineers to examine through a chatbot.

Niobia is initially focused on film coating, electroplating and fluidized bed reaction processes. Manufacturing errors can leave engineers spending days investigating a fix before production resumes, making faster analysis important for battery and chip producers.

The company’s co-founders, Gaurav Jha and Harveen Kaur, hold doctoral degrees in chemistry and have experience working at chipmakers and battery manufacturers. Jha said Niobia is pursuing a specialized market that includes some of the largest producers in both industries.

“It’s extremely niche; maybe people will never know about it,” Jha said. “But it’s like behind the scenes, this is powering everything… we just don’t see it in the regular world.”

Niobia uses open-source AI models trained further on battery, chip and research data. The approach is designed for customers that do not want proprietary information sent to a public AI cloud.

Customers can also retain their own encryption keys, keeping Niobia from decrypting their data. The company plans to make its technology more autonomous over time, with the eventual goal of supporting a fully self-driving factory for customers.