Glimpse, a startup focused on industrial quality control, has introduced a new product called Explore designed to significantly accelerate computerized tomography (CT) scanning for hardware manufacturers. The software leverages deep learning to process scans 10 to 30 times faster than traditional methods, addressing the slow throughput that often prevents companies from catching critical defects.
The impetus for such speed improvements is evident in recent industry history. Eric Moch, co-founder and CEO of Glimpse, cited the recall of Chevrolet Bolt vehicles from 2017 to 2022, which stemmed from improperly installed parts in LG battery cells. The incident ultimately cost LG $1.9 billion. Moch noted that major quality issues across the EV sector, affecting manufacturers like Volkswagen, Stellantis, Hyundai, and Toyota, could have been identified earlier if CT scanning had been feasible at scale due to current speed limitations.
Explore expands Glimpse’s capabilities beyond battery manufacturing to include aerospace, automotive, electronics, and medical devices. The software compresses and cleans scan data using algorithms that perform heavy processing on edge computers located at customer sites before transmitting results to the cloud. This approach allows for clear images to load in seconds within a web browser, enabling real-time defect detection.
According to Moch, the platform highlights deviations in specific measurements, such as layer distances within battery cells, flagging anomalies for review. The system also learns from customer feedback, retraining its algorithms to improve accuracy over time. Glimpse currently serves over 100 clients, including Anker, Lucid, and the U.S. Navy, generating mid-seven-figure annual revenue.
For companies lacking their own CT equipment, Glimpse offers a Scan On Demand service. The startup operates a facility in the Boston area and is opening a second center in the San Francisco Bay Area. Moch described this service as a potentially larger market segment, allowing businesses to pay modest fees for scanning rather than investing in expensive hardware upfront.
I’m skeptical about edge computing reliability. Can these algorithms really handle the noise in real factory environments without constant recalibration?
That Bolt recall is a stark reminder of why speed matters. catching defects early could save billions and prevent safety disasters.