
A bank may want to use AI, but can it trust the technology with sensitive customer data?
Bengaluru-based Soket AI is working to address that challenge with indigenous language models.
The startup has now secured nearly $4 million from Bajaj Finserv and Pennant Technologies to advance its AI research and explore enterprise applications. Pennant Technologies, an investee company of Bajaj Finance, has put in another $2 million.
The investment will support Soket AI’s work on large language models (LLMs) focused on coding, mathematics, and reasoning. Backed by the Government of India’s IndiaAI Mission, the company has launched EKA-1, a 26-billion-parameter foundation model under Project EKA.
Building AI for regulated industries
The partnership comes as financial services companies explore AI for lending, customer service, risk assessment and cybersecurity. However, using these systems also raises questions around sensitive customer data, security and control over technology infrastructure.
Bajaj Finance wants to explore models that deliver useful results at sustainable costs while letting the company run AI systems within its own infrastructure.
“Soket AI's efficient reasoning-led models can make advanced AI economically viable for financial services, while allowing us to deploy it securely within our own infrastructure, with full control over data,” said Rajeev Jain, Vice-Chairman and Managing Director of Bajaj Finance.
Sanjiv Bajaj also said the investment was part of the group's effort to strengthen India's sovereign AI capabilities and apply advanced research to real-world business challenges.
Pennant Technologies, which builds AI-powered lending solutions for financial institutions, plans to expand its work on AI agents and sovereign language models. Such systems could allow regulated institutions to retain control over their data and operational decisions.
From EKA-1 to larger models
Project EKA is intended to go beyond the current model. Soket AI aims to build trillion-parameter models supporting 100+ languages, including Indian and Global South languages, and programming languages.
The company is also focusing on making model training and deployment more cost-efficient. Its systems are being developed to run on organisations' own infrastructure, including on-premises environments.
Alongside EKA-1, Soket AI has launched Loop, an open-source framework for running long-running AI agent tasks.
Bajaj Finserv Intelligence follows a longer-term approach to applied research, combining startup investments, academic collaborations and in-house specialists. The initiative has a five-to-10-year horizon for developing scalable technology solutions.
The partnership gives Soket AI access to financial services expertise as it develops its models, while giving the Bajaj group an opportunity to explore how indigenous AI can be deployed in a sector where data security and reliability are essential.

