Niobia puts AI agents on factory lines to explain chip and battery defects (interview)

AI defect-detection and root-cause software for battery and semiconductor manufacturing lines.

That manufacturers will put self-contained AI agents directly on production equipment — keeping data out of the cloud — to cut defect root-causing from days to minutes.

Gaurav Jha and Harveen Kaur, both chemistry PhDs who worked at chipmakers and battery manufacturers, founded the San Francisco company around the expensive, slow defect investigations they had lived through.

In a pre-Disrupt call, co-founder Gaurav Jha says the agents run on self-contained GPUs mounted on production equipment like sensors, spotting pinhole-sized flaws and pooling data so engineers can query a chatbot for causes in minutes.

Engineers at chip and battery plants spend days or longer diagnosing failures before production restarts, and each flaw can cost millions.

Pinhole-sized defects such as scratches evade conventional inspection, leaving root-cause analysis to slow manual work.

Customers' refusal to move proprietary data into cloud AI forced the architecture: on-device GPUs, open-source models and customer-held encryption keys.

In industrial AI the blocker is not model quality but data custody: self-contained hardware and customer-held encryption keys are the price of admission where process data is the crown jewels.

As of October 13, 2026, Niobia is presenting at TechCrunch Disrupt 2026 as a Battlefield finalist; no customer names, revenue figures or funding rounds have been disclosed, and the near-term focus is the three specialty processes plus making the technology more autonomous.