Mihup & Qualcomm Partner to Launch On-Device Voice AI for India’s BFSI Sector

Indian Voice AI firm Mihup on Tuesday announced a collaboration with Qualcomm Technologies to develop and commercialise multilingual, enterprise-grade Voice AI solutions for the banking, financial services and insurance (BFSI) sector, designed to run directly on devices instead of relying heavily on cloud infrastructure.

The partnership aims to enable secure, scalable and high-performance Voice AI by shifting core intelligence onto devices powered by Qualcomm’s AI platforms. This on-device approach is expected to reduce latency, strengthen data privacy, lower bandwidth usage and improve reliability for enterprises operating in regulated environments such as financial services.

The companies said the collaboration comes as BFSI firms accelerate digital transformation and adopt AI-driven customer engagement tools, but face persistent challenges with cloud-only deployments, including high operating costs, connectivity constraints and concerns around sensitive customer data exposure. By running speech recognition and conversational intelligence locally, the solution is intended to improve customer service speed, agent productivity and governance of critical data.

“In a linguistically diverse and voice-first market like India, enterprises need secure, scalable AI to drive efficiency and financial inclusion,” said Savi Soin, Senior Vice President and India President at Qualcomm Technologies. He added that enabling Mihup’s vernacular Voice AI on Qualcomm’s on-device AI platforms would help deliver real-time conversational intelligence while supporting scalability and security.

Tapan Barman, CEO of Mihup, said combining Qualcomm’s neural processing unit (NPU) capabilities with Mihup’s proprietary voice technology eliminates recurring infrastructure costs and ensures full data sovereignty. He said the approach could deliver up to 80% cost reduction while enabling secure, real-time Voice AI for frontline operations.

Mihup’s contact centre solution will use a hybrid architecture, with high-volume workloads such as speech-to-text transcription and agent assistance running directly on devices, while more advanced processing can scale through the cloud when required. According to Mihup’s internal analysis, moving speech workloads on-device can reduce total cost of ownership by up to 78%, depending on deployment scale.

The companies said initial commercial deployments in India will serve as proof of scale, with plans to expand the solution to other developing markets and regulated industries globally, positioning on-device AI as a viable enterprise-grade alternative to cloud-dependent Voice AI systems.