Shunya Labs has announced the launch of Zero Codeswitch, a speech recognition foundation model designed to understand naturally code-mixed and multilingual Indian speech, a key limitation in existing voice-based AI systems.
The Gurugram-based voice AI company said the model is built to recognise how people in India actually speak, frequently blending Hindi, English and regional languages within the same sentence without relying on intermediate translation layers. The company says that Zero Codeswitch has achieved a 3.10% Word Error Rate (WER) on the OpenASR leaderboard, representing a 48% improvement over the next-best competing model, according to the company.
Shunya Labs said the model is designed to run efficiently on standard CPUs, reducing deployment costs by up to 20 times while maintaining sub-100 millisecond latency for real-time applications such as customer support, voice assistants and automated call centres.
“Shunya Labs was built with a focus on deep research rather than short-term marketing narratives,” Ritu Mehrotra, CEO and co-founder of Shunya Labs said. “With Zero Codeswitch, we are building foundational technology for Indian languages that prioritises accuracy, latency and real-world usability. Our goal is not just to adopt AI, but to build it at the foundation level in India.”
Unlike global speech models that are primarily trained on English data and later adapted for Indian languages, Shunya Labs said its foundation models are trained from the ground up on millions of hours of real-world Indian speech data. This includes variations in accent, dialect, pronunciation and slang across regions, allowing the system to better handle Hinglish and other code-mixed speech patterns.
“‘Shunya’ represents our philosophy of starting from first principles,” Sourav Bandyopadhyay, CTO and co-founder of Shunya Labs said. “With Zero Codeswitch, we are creating an intelligence layer that truly listens, engineered for India’s linguistic diversity and optimised for real-world deployment,.”
The company said Zero Codeswitch is intended for enterprise and public-sector use cases where data privacy is critical. The model can be deployed on-premises or in air-gapped environments, allowing organisations to train domain-specific versions while retaining control over sensitive data. Shunya Labs said it complies with HIPAA, SOC 2 Type II and ISO 27001 standards and supports CPU-first deployments to reduce reliance on specialised GPU infrastructure.
The post Shunya Labs Unveils AI Model for India’s Code-Mixed Speech appeared first on Analytics India Magazine.
