ShepHertz Unveils On-Device AI Models for Offline Use
ShepHertz has introduced three AI model families—Tatva, Sanjaya, and Drashta—designed to run on devices with limited computing resources and unreliable internet access. They target command processing, telemetry monitoring, and machine vision. The company is also opening its AI trust toolchain, beginning with Shuddhi, while independent testing and broader availability will determine their practical impact.
ShepHertz, an Indian technology firm, has announced Tatva, Sanjaya, and Drashta as model families for on-device artificial intelligence. They are meant for hardware with limited compute and no dependable internet link. The company presented them at Drone Expo 2026 in New Delhi under its AgentAnywhere platform, citing possible uses in satellites, drones, smart glasses, and smart rings.
Tatva converts ordinary-language directions into commands a machine already understands, covering English, Hinglish, and Indian languages while rejecting requests beyond its allowed set. Sanjaya watches telemetry streams, and Drashta addresses machine vision. ShepHertz is also opening a trust-focused tool set for AI, starting with Shuddhi, and highlights human sign-off plus records that can be traced. Tatva Edge remains in preview.
If these models work as described, people and organizations operating in remote or bandwidth-limited settings could gain more responsive tools that keep working without cloud links. Industrial crews, drone operators, satellite users, and wearables owners may benefit, while reliance on local decisions could raise questions about oversight. Independent testing and wider release may shape whether benefits outweigh limits.