In June 2023, the Monetary Authority of Singapore (MAS) launched the 8th edition of the Global FinTech Hackcelerator, titled “AI in Finance Global Challenge”. The competition was conducted in partnership with AI Singapore (AISG) and powered by Oliver Wyman. It seeks to produce innovative and market-ready AI solutions that can transform the financial services industry.

Pints AI emerged as one of the champions of this Challenge at the Singapore FinTech Festivals Awards and were subsequently awarded the AI in Finance Global Challenge (AIGC) Startup Grant from AI Singapore.

We recently caught up with Calvin Tan, CTO and Partha Rao, CEO of Pints AI of to find out more….

Congratulations for being awarded the AI Singapore Startup Grant in the recent AI in Finance Global Challenge, organised by MAS and AISG.

Q1: Tell us more about your company

Partha: Pints AI is a privacy-first enterprise Gen AI platform that allows clients in the financial industry to use all their data, including the confidential & proprietary data and build secure AI Agents. We combine our proprietary private Compact Language Models and the Pints AI Orchestration Framework enabling insurers, asset management companies and banks to build secure AI tools and apps, without having to worry about their data travelling out of their secure infrastructure.

We started working on building an AI Agent that can help finance professionals and retail investors learn, investigate and discover new investing opportunities. We discovered that the core problems while deploying Gen AI solutions were mainly two-folds:

1. Concerns over the cost of compute:  Compute is still a large factor in decision making, and organisations have to balance the cost of AI deployment with presently uncertain ROI. This stops widespread deployment, and many companies are limiting Gen AI tools to POC or allow only limited number of employees to use Gen AI.
2. Privacy of data: Given that all current feasible GenAI providers run on public cloud, it is not possible to use such solutions for sensitive data.

This restricted adoption of Gen AI to the most simplistic use cases – e.g. marketing, customer service using publicly available data like marketing material or product information.  Financial Institutions and Regulators in multiple countries are concerned about the use of Gen AI solutions involving sensitive information or PI data. Lack of clear policy guidelines limit the organisations from using their most strategic data with Gen AI solutions. With our solutions, such barriers do not exist as we deploy compact models directly into the secured environment that is fully controlled by the client’s infrastructure teams – whether it is on a private cloud or even on premise.

Q2: Can you share with us the motivation behind joining the Challenge

Calvin: We wanted to leverage the challenge as a platform to innovate and develop AI solutions tailored for the Asian market. The opportunity to present to a large audience, at Singapore Fintech Festival was compelling – and the chance to interact with the team from AI Singapore was a great advantage.

Q3: Can you share more about your winning proposal

Calvin: Our proposal was focused on developing a new generation of AI solutions specifically tailored for the financial sector which will provide customisable, powerful, and cost-effective AI solutions. This involved creating compact language models that are significantly less resource-intensive compared to current large models like GPT-4, making them ideal for deployment in environments with stringent privacy requirements or limited hardware capabilities. This solution will help increase privacy and efficiency, reduce costs, and enhance accessibility to powerful AI tools for businesses. Since the win, we have pushed ahead with our research and our compact model 1.5 Pints will be out soon. We plan to continue down this path and develop the Pints MOE – a framework that will allow financial organizations to unlock all of their organizations data, without worrying about its misuse.

Q4: What is/are key to bringing this solution into fruition? What milestones do you hope to achieve with this funding?

Calvin: To bring our solution to fruition and achieve our goals with the funding, we are committed to developing compact, efficient AI models tailored for private use, including the ones that can be deployed on personal devices. This approach diverges from the prevalent trend of creating larger models, focusing instead on sub-10 billion parameter models that are more suited for the diverse and unique requirements of the Asian market. Our key objective is to develop a language model architecture that is not only private and customisable but also powerful and cost-efficient, ensuring accessibility and practicality for a broader audience.

Through this funding, we aim to achieve significant advancements in research and product development. We will be able to advance our research into next-generation NLP architecture, moving beyond the Transformer architecture. Additionally, we will acquire GPUs to alleviate the high costs associated with cloud computing, co-fund and co-research projects with SUTD, and hire AI talent. This funding will also support the development of economically viable AI products that can operate independently on personal devices without compromising performance. In the near future, we aim to expand our reach within Asia, leveraging our innovations to provide industry-grade AI tools to a broader audience, thus enhancing both social and economic outcomes regionally.

Q5: Who are the people involved and how?

Partha: We are a 7-member team, including Calvin and I. Besides the core team, we have interns, round the year, who support our team in developing our models and architecture. In addition, we have a collaboration with the Singapore University of Technology and Design (SUTD) for research development.

Q6: Were there any challenges along the way and how did you work around it? What are the biggest challenges you anticipate in the next 6-12 months?

Calvin: Originally, we faced uncertainty around the fundraising environment. There was a general view that building LLMs was best left for US and Chinese big tech, and we were asked why we were pursuing this at all. But our clients were telling us otherwise – they wanted private and controlled deployment of Gen AI – so we decided to focus on growing our revenue and proving the demand. We also realised that targeting large financial institutions in Singapore may not be feasible, as the competition here was intense and hence decided to expand across SEA.

We will soon release the 1.5 Pints, our high-performance compact language model, and start work on Pints MOE – a mixture of experts framework that will be able to tackle the complexities of language, context and industry nuance. In addition, we are launching Pints RevSurge, a platform that leverages our private architecture to boost financial RM & Advisor productivity. Also, we will expand out of Asia and are pursuing partnerships with System Integrators and consulting companies in Australia and Europe.

Q7: How do you plan to scale your team in line with your growth objectives?

Partha: We plan to scale 3 key functions of our team: (1) research, (2) product development, and (3) sales. For research, it will help us unlock AI frontiers that aligns with the product requirements, to significantly enhance the quality of our offerings, and sharpen our competitive edge. For product development, having the best technology is only as good as translating it into useful and delightful products that improve lives, and users want more of it. Finally, a product will need sales personnel to put it out there and make users know the benefits and make that purchase because he/she is convinced that it is worth it.

Q8: How do you differentiate yourself from the other market players out there?

Calvin: Pints AI specialises in compact, on-premise AI solutions tailored for specific industries, distinguishing itself from larger tech firms, like ChatGPT, that are focused on general applications. Unlike larger tech companies that offer generalised AI tools, Pints.ai focuses specifically on the financial sector, providing bespoke solutions that are finely tuned to the industry’s unique needs. By focusing on developing smaller, yet still powerful models, Pints AI offers a practical alternative to the large, cumbersome models that dominate the market. This approach not only reduces the hardware footprint required but also aligns with trends towards more sustainable tech practices. Recognising the industry’s sensitivity to data privacy, we emphasise on-premise solutions that allow clients to maintain complete control over their data.

Q9: How do you plan to utilize the grant funding you’ve received?

Calvin: Given the current limited AI talent in Singapore, Pints AI will utilise the grant to help train AI Engineers. This has been a critical piece that enabled Pints AI to train a staff of 3 software engineers and 4 interns into AI practitioners, to improve our AI capability and continue to push frontiers. We will expand this with the help of the grant. Specifically, we will collaborate with SUTD to train and expose Computer Science undergraduates to AI/NLP, and put personnel up for training courses, such as those offered by AISG or General Assembly. We also want to organise “hacker café sessions” to share AI knowledge, to broaden interest in AI, and attract high quality talent.

Q10: What is next after this? What are your long-term vision and goals for the company?

Calvin: Our vision is to bring industry-grade AI tools to the wider public for social and economic benefits. Gen AI offers incredible benefits, but concerns around privacy, ownership of data, opaqueness of AI models, and cost of compute are restricting its wider deployment. We want to remove these concerns and democratize access to AI, while ensuring that our clients have full control over the entire product – from LLM to data to the output generated.

Based in Singapore, we are committed to positioning Singapore as a global hub for AI research and development. Our long-term goals include:

1. Make our AI tools accessible to a broader audience, including small and smaller FIs that may not have the resources to develop their own AI solutions.
2. Continue to prioritize the privacy and security of our clients’ data, ensuring that our solutions comply with the highest standards of data protection and privacy regulations globally.
3. Invest in research and development to stay at the forefront of AI innovation, including adapting Pints AI framework to work on your personal devices.

We are dedicated to putting Singapore on the global map for AI research and development. Singapore offers some unique advantages – a strategic position, robust infrastructure, stable forward-looking policy, and a commitment to innovation. It also helps that we are a financial hub! We are hoping we can leverage these to and position Pints AI and Singapore as a leading center for AI excellence and innovation.