NPCI teams up with Nvidia to build AI layer for payments: What it means

NPCI says the effort builds on its FiMI model and UPI Help Assistant, and is aimed at creating a payments-native AI foundation aligned with India's data sovereignty and regulatory needs

At the India AI Impact Summit 2026, the National Payments Corporation of India (NPCI) has outlined a broader push to build what it describes as a payments-native, sovereign AI layer for India’s digital payments infrastructure. The organisation announced a collaboration with Nvidia on February 18 to scale its in-house AI work, building on the recent launch of FiMI (Finance Model for India), a domain-specific language model developed for India’s payments ecosystem.

 

According to NPCI, it plans to use Nvidia’s NeMotron family of open models as part of its model development work to create a payments-native AI foundation model that aligns with India’s regulatory and data sovereignty requirements. 

 

From FiMI to a broader AI layer

The announcement follows NPCI’s release of FiMI a day earlier at the India AI Impact Summit. FiMI is a domain-specific language model built in-house by NPCI and is currently used to power the UPI Help Assistant, a pilot AI-based support system for UPI users.

 

NPCI has said FiMI was developed to address the limits of general-purpose language models in high-scale, high-trust payment environments. The model has been trained and fine-tuned on Indian financial and synthetically generated payments data, and is designed to handle tasks such as transaction dispute queries, mandate lifecycle management, and regulatory or ecosystem-related questions.

The UPI Help Assistant, which runs on FiMI, is already deployed at a national scale as a pilot. It supports English, Hindi, Telugu and Bengali, with more Indian languages expected to be added over the next six to eight months. NPCI has also published a technical paper detailing how the model was built and evaluated.

 

In its statement, NPCI said the next phase of its AI work will move beyond use-case–specific agents, like the Help Assistant, towards a more general, scalable AI layer for the payments ecosystem. This includes exploring architectures such as Mixture of Experts to support high-volume, low-latency environments, and gradually expanding multilingual and agent-oriented capabilities.

Why NPCI is building this in-house

In a conversation with Business Standard at the India AI Impact Summit, NPCI Executive Director-Growth, Sohini Rajola, said the organisation’s AI strategy is built around “three pillars”: internal use to improve efficiency, building models that the wider payments ecosystem can build on, and ensuring that AI systems used in payments reflect Indian data, languages and context.

 

Rajola also said the starting point for NPCI’s AI deployments has been trust and customer experience, including giving users better visibility and control over mandates, and improving how grievance redressal is handled at scale.

 

Responding to the question about the models that are currently in use, Rajola said that the models were “100 per cent developed by NPCI.” She added, “We obviously work on open source platforms but it is developed in-house by our AI scientists and we train it on Indian data and the models are refined and fine-tuned absolutely natively.”

What Nvidia brings to the partnership

As part of the collaboration, NPCI said it will use Nvidia’s Nemotron models and the NeMo framework in its model development work. Nvidia has said NPCI is exploring training FiMI using the Nemotron 3 Nano model along with its own datasets, with the aim of supporting multilingual customer service across India’s banking ecosystem.

 

Vishal Kanvaty, Chief Technology Officer at NPCI, said, “through this collaboration with NVIDIA, NPCI aims to advance AI capabilities designed specifically for India’s payments ecosystem. Drawing from our experience of operating population-scale, real-time payment systems, this initiative is designed to create a sovereign, payments-native AI foundation that strengthens trust, resilience, and security, while remaining aligned with India’s regulatory and data sovereignty requirements. As we evolve from use-case–led AI deployments to a foundational and scalable AI layer, our focus remains on enabling the broader ecosystem to innovate responsibly through robust governance frameworks and secure, future-ready infrastructure.”