NimbleEdge has already deployed its platform on over 30 million devices across sectors like gaming and e-commerce, demonstrating its scalability and effectiveness in real-world applications.

An Agent Marketplace has been introduced, making it easy for developers to discover and integrate pre-built AI agents into their mobile applications.

The platform supports building privacy-preserving, scalable AI solutions that align with India’s goal of creating trusted AI infrastructure for 1.4 billion people, enabling AI to run directly on smartphones.

NimbleEdge has researched techniques like contextual sparsity and developed custom C++ kernels to enhance generation speed and reduce memory usage, ensuring efficient performance on resource-constrained devices.

Experts and industry leaders recognize on-device AI as a major advancement for making intelligent applications more accessible, private, and performant, with Columbia University’s Vishal Misra emphasizing its significance.

NimbleEdge has launched 'DeliteAI', an open-source, on-device AI platform tailored for mobile devices, enabling developers to create private, scalable AI experiences without relying on the cloud.

DeliteAI has been reported to run on over 30 million devices, including smartphones as old as eight years, with low battery consumption compared to standard API calls.

NimbleEdge's CTO highlighted that DeliteAI provides the infrastructure for developers to deploy private AI agents at scale, aligning with India's extensive smartphone user base and national AI mission.

Developers can bring their own models and orchestrate AI workflows locally, supporting diverse hardware and runtimes like ONNX, LiteRT, and ExecuTorch for fully offline, customized AI solutions.

The DeliteAI platform and its components, including source code, are now available on GitHub, along with a community Discord, to democratize access to advanced AI on billions of devices.

DeliteAI includes a production SDK with an optimized inference stack, an on-device Python runtime for AI workflow orchestration, an Agent Marketplace for pre-built AI agents, and a fully on-device conversational AI assistant aimed at enhancing user productivity and privacy.

The platform addresses common cloud AI limitations such as latency, privacy risks, and operational costs, by allowing models like Llama, Gemma, or Qwen to run inference locally on devices, ensuring data privacy.

The platform integrates with various open models and features an Agent Marketplace that offers reusable on-device AI agents for tasks such as text summarization, voice interaction, and recommendations.