RESEARCH PARTNERSHIP
IDFC FIRST Bank and Sarvam partner to build a self-improving bank
The joint R&D Lab will work across frontier research, post-training and AI safety to build banking systems that can learn from outcomes and improve over time.
IDFC FIRST Bank and Sarvam are establishing a joint R&D Lab to explore what it means to build a bank that can continuously improve how it uses AI.
AI capabilities are advancing quickly, but making those capabilities useful in banking requires more than access to better models. Systems need to understand the bank, operate within its controls, learn from product outcomes and improve without requiring the institution to rebuild around every change in the underlying technology.
The Lab will bring together IDFC FIRST Bank’s banking expertise and Sarvam’s research and engineering to work on this problem.
Building a self-improving bank
A bank produces a large amount of useful signal through everyday operation, including customer interactions, decisions, human corrections, exceptions and eventual outcomes. A self-improving system should be able to learn from this experience, evaluate what worked and use that signal to improve the next version.
Over time, this creates a different relationship with AI. New models and capabilities can continue to arrive, while the bank retains what it has learned about its customers, processes, policies and decisions.
As frontier capabilities become broadly available, the ability to learn from them and adapt them to the institution may become increasingly important.
The joint R&D Lab will work on the research and systems required to make this possible in a regulated banking environment.
The Lab will bring research much closer to the bank’s day-to-day operations. New models and techniques can be tested against IDFC FIRST Bank’s own problems, evaluated in production and improved using the outcomes they produce.
This gives the bank a way to keep adopting better AI as the technology advances, without rebuilding its systems every time.
The Lab will also create an environment for researchers and engineers to work on fundamental AI problems in the context of financial services, where questions of reliability, security, explainability and control are as important as capability.
The foundations of a self-improving bank
The R&D Lab will focus on the research, systems and safeguards required for AI to learn continuously while operating within the controls of a regulated bank.
Frontier research
Advancing AI and large language models on problems relevant to banking and financial services, and exploring new capabilities as the underlying technology evolves.
Post-training factory
Building the infrastructure to evaluate, adapt and improve models using banking tasks, human feedback and real-world outcomes.
The objective is to turn production experience into a continuous source of learning, so systems become more capable and more specific to the bank over time.
Regulator-grade safety and AI security
Developing evaluation, governance, security and safety systems alongside model capability.
AI systems operating inside a bank must remain observable, auditable and controllable as they become more capable. The Lab will work on the technical foundations required to make continuous improvement compatible with the standards expected of a regulated financial institution.
Building for what comes next
Models will improve, costs will change and entirely new ways of using AI will emerge.
The more durable capability may therefore be the ability of the institution itself to keep learning, to evaluate what is new, absorb what is useful and carry forward what it has already learned.
The ambition is to build a bank that keeps getting better as AI gets better.