Pints AI Wins MAS-AISG Startup Grant for Privacy-First Finance LLMs
Source: AI Singapore
Pints AI, a Singapore-based startup building compact on-premise language models for financial services, has been awarded the AI in Finance Global Challenge Startup Grant by AI Singapore and the Monetary Authority of Singapore. The grant will fund development of sub-10 billion parameter models deployable entirely within clients' secure infrastructure, addressing the growing tension between AI adoption and data privacy in banking and insurance.
A Singapore startup taking a contrarian bet on small, private AI models for finance has been awarded the AI in Finance Global Challenge Startup Grant from AI Singapore and the Monetary Authority of Singapore. Pints AI, a seven-person team with a collaboration with the Singapore University of Technology and Design, won the grant after emerging as a champion in MAS's 8th Global FinTech Hackcelerator — a competition that specifically sought market-ready AI solutions for the financial sector.
The company's pitch stood out because it goes against the prevailing current of ever-larger language models. Pints AI builds compact models under 10 billion parameters that deploy entirely within a client's own infrastructure — no cloud handoff, no data leaving the building. Its co-founders, CEO Partha Rao and CTO Calvin Tan, argue that most financial institutions are stuck in a GenAI pilot loop precisely because they can't risk sending sensitive data through public cloud APIs. Their solution: drop a small, capable model directly into the bank's secure environment, train it on internal data, and keep the whole stack behind the firewall.
The timing works in their favour. Singapore's financial regulators have been signalling discomfort with how GenAI models handle proprietary data, and global banks are increasingly cautious about the AI supply chain. Pints AI's approach — compact models purpose-built for finance, running on-premise or in a private cloud — sidesteps the privacy, cost, and latency objections that have kept enterprise AI adoption in pilot purgatory. The grant will fund GPU acquisitions, co-research with SUTD, and a structured programme to train software engineers into AI practitioners, addressing the talent bottleneck that Rao and Tan say is their binding constraint.
Why it matters for Singapore: MAS and AISG structured this grant programme to surface homegrown solutions to problems that big-tech AI products aren't solving. Pints AI's bet on compact, finance-specific models is exactly the kind of differentiated approach Singapore's AI ecosystem needs — it doesn't try to out-LLM OpenAI but instead solves a practical, high-value problem that global banks face everywhere. If it works, it gives Singapore a genuinely differentiated play in the enterprise AI market: not the biggest models, but the most trusted ones.