Indian voice AI company Mihup has entered into a collaboration with Qualcomm Technologies to co-develop and commercialise multilingual, enterprise-grade voice AI solutions for the banking, financial services and insurance (BFSI) sector, optimised for on-device AI platforms.
The partnership focuses on enabling voice AI capabilities that run directly on Qualcomm-powered AI PCs and devices, reducing reliance on cloud infrastructure while delivering low latency, stronger data privacy and improved cost efficiency.
The companies said the approach is particularly suited for regulated industries such as BFSI, where governance, security and real-time performance are critical.
As financial institutions accelerate digital transformation, cloud-only voice AI deployments continue to face challenges related to bandwidth costs, latency and exposure of sensitive customer data.
The Mihup-Qualcomm collaboration aims to address these gaps by shifting high-volume speech workloads to on-device neural processing units (NPUs), enabling faster customer engagement and improved agent productivity.
“In a linguistically diverse market and voice-first like India, enterprises and industries such as BFSI need secure, scalable AI to drive efficiency and financial inclusion,” Savi Soin, Senior Vice President and India President, Qualcomm Technologies, said. “Our collaboration with Mihup demonstrates how on-device AI on Qualcomm platforms can bring real-time, multilingual conversational intelligence closer to users. By enabling Mihup’s vernacular Voice AI to run directly on our on-device AI platforms, we are helping deliver high-performance, secure solutions for frontline systems, thereby supporting broader financial inclusion and global scalability.”
Mihup said its voice AI contact centre solution will follow a hybrid architecture, with vernacular speech recognition and custom language models optimised to run locally on Qualcomm NPUs. This allows real-time transcription and agent assistance to be handled on-device, while more complex workloads can scale to the cloud as required.
“In a market as vast as India, it requires intelligence that is both local and affordable. By merging Qualcomm’s NPU capabilities with Mihup’s proprietary voice stack—engineered specifically for vernacular and on-device usage—we have eliminated recurring infrastructure costs. We are delivering an 80% cost reduction and total data sovereignty directly on the silicon. This is the shift from vision to production reality, providing the Indian frontline with voice AI that is secure, instant and built to scale,” Tapan Barman, CEO, Mihup, said.
According to Mihup’s internal deployment analysis, moving high-volume speech-to-text and real-time agent support workloads on-device can reduce total cost of ownership by up to 78%, depending on scale and usage.
For BFSI institutions, the solution enables multilingual customer service, sales and collections by mirroring how customers naturally speak, while delivering faster responses and more consistent experiences.
With commercial deployments underway in India, the companies said the collaboration will serve as a proof-of-scale for enterprise adoption of on-device voice AI. While designed for India’s linguistic diversity and cost sensitivity, the solution is expected to be positioned for expansion into other developing markets and regulated industries globally.
Your reaction