U.S. startup Niobia is developing an AI system that helps battery and semiconductor manufacturers detect defects and determine why they occurred. The software analyzes data from production equipment, identifies even microscopic damage, and allows engineers to ask a chatbot questions to find the root cause of a malfunction more quickly.

Instead of spending days investigating a production failure, specialists will be able to get guidance in a matter of minutes. This is especially important for industries where a single mistake can lead to multimillion-dollar losses and delay product launches, TechCrunch reports.

AI searches for defects directly on the production line

Niobia installs AI agents directly on production equipment. They run on compact graphics processors placed next to the machines and serve as a kind of intelligent sensors. The system is capable of spotting even the tiniest defects, such as scratches and holes the size of a pinhead.

But detecting a problem is only the first stage. The software gathers production-process data in one place so engineers can investigate possible causes of a malfunction with the help of a chatbot. Specialists ask questions, clarify details, and analyze the collected information in an attempt to determine at what stage the error occurred and how to eliminate it.

This approach is intended to reduce the time between detecting a defect and resuming production. In the development of new batteries and microchips, technological processes are constantly changing, so disruptions are inevitable. While engineers are figuring out what went wrong, equipment may sit idle and product release schedules may slip.

Why manufacturers need such systems

Niobia is targeting two industries where the cost of mistakes is especially high: semiconductor and battery manufacturing. Their products are essential for electronics, transportation, and many other technologies, while the manufacturing processes themselves are becoming increasingly complex.

The startup’s co-founders, Gaurav Jha and Harvyn Kaur, hold PhDs in chemistry and previously worked at companies that manufacture microchips and batteries. According to Jha, this experience helped them see the problem from the inside: engineers have to sort through enormous amounts of data to determine the cause of even a minor deviation.

Niobia is betting on a highly specialized market that may go unnoticed by ordinary consumers, even though the related processes are used to manufacture a wide range of products — from headphones and glasses to shoes and wallets.

The company became a finalist in the Startup Battlefield competition at the TechCrunch Disrupt 2026 conference in San Francisco.

Why Niobia uses open AI models

The system is based on open-source models that Niobia further trains on research data related to batteries, semiconductors, and manufacturing processes. This tuning is meant to help the algorithms recognize specific defects and better understand the context in which they arise.

However, for potential customers, not only the accuracy of the analysis matters, but also information security. Production data may contain trade secrets and information about technologies that companies do not want to hand over to third-party providers of cloud AI services.

According to Jha, customers are not willing to send their own data to the public cloud, even if the provider assures them it is safe. That is why Niobia uses on-premises computing and allows customers to manage encryption keys themselves. In this configuration, the company cannot decrypt a client’s data without the corresponding keys.

This is especially important for battery and chip manufacturers operating under intense technological competition. Protecting intellectual property here may be no less important than the speed of defect detection.

From error detection to a fully automated factory

At the first stage, Niobia will focus on three manufacturing processes: thin-film coating, electroplating, and fluidized-bed reactions. The latter method involves a flow of gas or liquid keeping solid particles in suspension, making it possible to carry out chemical processes involving them.

Choosing narrow specializations will allow the company to adapt AI to specific manufacturing tasks rather than trying to cover every type of industrial equipment at once.

In the future, the startup wants to make the system more autonomous. The ultimate goal is a factory that can independently monitor processes, detect deviations, and respond to problems with minimal human involvement.

For now, Niobia offers a tool to assist engineers rather than fully autonomous manufacturing. To move closer to its stated goal, the system will need not only to detect defects, but also to reliably determine their causes, propose safe adjustments, and prove its effectiveness on real production lines.