Indian technology company ShepHertz has introduced three AI model families designed to run on devices with computing resources, including systems that cannot rely on a constant internet connection. This announcement addresses a key AI challenge: making it useful when cloud access is unavailable, expensive, or unsuitable.

Much of today’s AI experience depends on sending requests to servers over the internet.

That approach works well in many environments, but it can limit equipment operating in remote locations, industrial sites, or other places where connectivity is unreliable.

ShepHertz says its new model families, Tatva Sanjaya and Drashta, are designed to bring AI to the machines they run on.

The company unveiled them at Drone Expo 2026 in New Delhi as part of its AgentAnywhere platform. The company says it built the models in India and can run them offline on constrained hardware, with applications ranging from satellites and drones to smart glasses and smart rings.

Each model family has a purpose. Tatva translates plain-language instructions into commands a machine already recognizes.

The company says it can handle English, Hinglish, and Indian languages, but it refuses instructions outside the permitted command set. Tatva Edge is currently in preview, so its broader availability and real-world performance remain to be established.

Sanjaya focuses on system telemetry, the streams of measurements and operating data equipment generates.

It is intended to flag patterns for operators to review, with potential uses in infrastructure, industrial systems, and other settings where early warning is critical. ShepHertz says Sanjaya is working with design partners.

Source link