Shiprocket launches Shunya.ai, a sovereign multimodal AI model for MSMEs
Trained on Indian voice and text data, including Shiprocket’s years of customer support call logs, Shunya.ai is designed for offline and low-power deployments, with a lightweight version planned to run on CPUs.
Logistics and ecommerce enablement firm Shiprocket has launched Shunya.ai, a sovereign, multimodal AI model developed entirely in India to serve the country’s micro, small and medium enterprises (MSMEs) and digital commerce ecosystem.
The model, which supports voice, text and image inputs across more than nine Indian languages, is being positioned as a foundational AI layer for Bharat-focused businesses.
There are two versions of the models: one with 15 billion parameters and another with 45 million parameters. The company claims that the model's performance puts it among the top 5 LLMs available in the market.
Shunya.ai was built independently from global large language models such as OpenAI’s GPT or Meta’s LLaMA. The technology was developed in partnership with US-based Ultrasafe Inc, with foundational model IP provided under an exclusive, perpetual license.
All India-specific deployments, including agentic improvements and fine-tuning, are owned and operated domestically.
"This is not just AI made in India, it's AI made for India,” said Saahil Goel, MD and CEO of Shiprocket. “We've adopted Shunya.ai from the ground up for Indian languages, commerce workflows and MSME needs... we're giving over 1.5 lakh sellers instant access to tools that are intelligent, focused, and scalable, levelling the playing field for businesses across Bharat."
Built on an AI-optimised GPU infrastructure hosted entirely within India, the platform was developed in collaboration with Larsen & Toubro’s Cloudfiniti unit, ensuring regulatory compliance and data localisation.
Seema Ambastha, CEO of L&T Cloudfiniti, said: “We are proud to power India’s first sovereign AI engine with secure, high-performance GPU infrastructure, fully hosted within the country. Our partnership with Shiprocket and Ultrasafe reflects our deep commitment to fostering secure, scalable AI innovation that drives India’s digital transformation."
The model has already been deployed internally at Shiprocket to assist with customer service and commerce automation tasks. For enterprises, Shunya.ai offers custom workflow digitisation and LLM development. MSMEs can access pre-built tools via a freemium pricing model starting at Rs 499 per month.
Raaid Hossain, CEO and co-founder of Ultrasafe, said: “This model is the first of its kind in India, not just in architecture, but on purpose... It’s multilingual, tailored to Indian use cases, and built from the ground up with trust, privacy, and scalability at its core.”
Trained on Indian voice and text data, including Shiprocket’s years of customer support call logs, Shunya.ai is designed for offline and low-power deployments, with a lightweight version planned to run on CPUs.
Beyond commerce, the company sees use cases in agriculture, legal services, logistics, and government. Early pilots have shown 30–40% reductions in time spent on catalogue and content creation, according to internal data.
Shunya’s development was entirely self-funded, with about $40 million invested by the founders, and without any external venture capital.
Edited by Jyoti Narayan