San Francisco-based startup Niobia is betting on a critical but often overlooked step in battery and chip manufacturing: helping engineers understand how and why problems occur. The company says its AI software can flag production defects and allow engineers to query the causes via a chatbot, aiming to reduce response times from days to minutes.
The general workflow is as follows: customers deploy AI agents on the production line to detect defects, after which human engineers determine potential causes of product failure. These agents run on dedicated GPUs directly installed on manufacturing equipment, functioning like sensors. They can identify flaws as small as pinholes, such as scratches. Niobia’s AI also centralizes all data in one location, enabling engineers to ask a chatbot where issues occurred and further investigate the root causes.
Speed is critical because chips and batteries are among the most important products globally—and they are becoming increasingly scarce. As manufacturers continuously develop and improve technology, issues during R&D are almost inevitable. Engineers often spend days or longer identifying solutions before production can resume. Each defect or mistake can be extremely costly—sometimes reaching millions of dollars—and significantly delay manufacturing.
Niobia’s co-founders, Gaurav Jha and Harveen Kaur, both hold Ph.D.s in chemistry and have worked at chip and battery manufacturers, giving them firsthand experience with industry challenges. In a call ahead of TechCrunch Disrupt 2026, Jha explained that although Niobia’s work is highly specialized, it serves a multi-billion-dollar market that includes some of the largest battery and semiconductor manufacturers today.
“This is very niche; maybe people will never know about it,” Jha said when discussing the battery and chip manufacturing markets the company serves. “But it’s like behind the scenes, powering everything… we just don’t see it in the real world.”
Jha added, "Every product around you—such as glasses, shoes, slippers, wallets, and headphones—involved in these processes touches every part of them."
Niobia’s agents run on open-source models, which the company further trains using data from battery and chip research and other sources. This training enables the models to detect pinhole-sized defects. Jha said Niobia primarily chose open-source models because customers do not want their highly sensitive corporate information entering public AI clouds, let alone being used to train models.
“No client wants their proprietary data to go into the cloud, even when we tell them it’s secure,” said Jha. Given the high-stakes competitive landscape in the battery and chip industries, where nations are vying for innovation, the startup also highlights its cybersecurity practices. For example, Niobia allows customers to hold their own encryption keys, so the company itself cannot decrypt customer data.
Jha said the company is currently focusing on three core areas: thin-film coating, electroplating (the process of depositing metal onto a surface using an electric current), and fluidized bed reaction (a chemical process in which solid particles are suspended in a gas or liquid).
He said the company also aims to make the technology more autonomous, with the ultimate goal of building a fully self-driving factory for customers.
If you'd like to learn more, Niobia will be presenting on stage at TechCrunch Disrupt in San Francisco this afternoon. Come see Niobia and the other 19 startups competing alongside it as part of the Startup Battlefield competition, running from today, Tuesday, through Thursday, October 15.