Telit Cinterion has revealed an edge AI SDK alongside planned AI-enabled cellular module variants designed to run machine learning on-device.
The SDK runs machine learning models directly on the cellular module’s internal hardware. By executing workloads natively, the architecture operates without an external AI accelerator or a dedicated companion processor.
Scheduled for release in the fourth quarter of 2026, the kit embeds LiteRT (formerly TensorFlow Lite) into the Linux-based firmware of the company’s upcoming hardware. The planned hardware rollout covers 4G, 5G RedCap, and high-performance 5G module variants.
Native runtime integration and model portability
LiteRT supports standard .tflite files, allowing engineering teams to train and refine machine learning models within established toolchains. Developers can load the finished file straight onto the cellular module without compiling for a proprietary runtime environment.
Models running on a standard personal computer or a Raspberry Pi can be transferred directly to the module firmware. The compact runtime operates within the memory and compute limits of lower-cost modules that cannot accommodate full-sized machine learning software stacks.
This footprint brings local inference capabilities to hardware categories historically restricted to wireless data transmission.
Marco Argenton, SVP of Product Management at Telit Cinterion, said: “Industrial IoT teams should not have to redesign their entire device architecture to add practical AI capabilities.
“By bringing a lightweight AI runtime into the cellular module, we help customers reduce system complexity and accelerate the path from proof of concept to a connected industrial solution that can operate reliably in the field.”
Pipeline architecture and thermal thresholds
To assist system integrators, the SDK includes sample code covering the entire inference pipeline. The software manages sensor data acquisition, runs local preprocessing routines, executes the model, and passes the output to the host industrial application. Integrators maintain complete control over the application logic, target models, and final deployment setup.
Internal proof-of-concept testing across computer vision workloads confirmed the processing overhead of the runtime. During image classification and object detection trials, inference tasks consumed a maximum of 17 percent of CPU capacity. That headroom keeps heat generation below the threshold where thermal throttling reduces cellular data rates across 4G and 5G connections.
Target industrial deployments highlighted for the technology focus on telemetry analysis, visual capture, and remote audio monitoring.
Integrators can process vibration and acoustic readings from industrial pumps, motors, and bearings to detect mechanical anomalies before failures develop. The runtime also processes sound patterns to detect broken glass or facility alarms at unstaffed installations, or links with local cameras to read legacy analogue meters without replacing infrastructure.
Vishal Batra, VP of Software Engineering at Telit Cinterion, explains: “The real engineering challenge isn’t simply running AI at the edge. It’s delivering secure, efficient, and reliable intelligence within the constraints of a connected embedded device.
“Our edge AI SDK enables developers to deploy standard AI models directly on Telit Cinterion modules, accelerating development without compromising the trusted connectivity industrial IoT applications require.”
Telit Cinterion plans to ship the edge AI SDK and the accompanying 4G, 5G RedCap, and high-performance 5G hardware variants in Q4 2026.
See also: USI launches edge AI camera for factory-floor inspection
Want to learn more about the IoT from industry leaders? Check out IoT Tech Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including AI & Big Data Expo and the Cyber Security Expo. Click here for more information.
IoT News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
