Indian banks want pre-trained LLMs to be deployed for them which is a challenge: Maveric Systems' CTO

Founded in 2000, Chennai-based Maveric Systems delivers BankTech solutions, helping financial institutions tackle compliance, digital operations, and customer experience challenges. Operating in 15 countries, the company aims to triple its revenue by 2025 through innovation and expansion.

In an interview with TechCircle, Kishan Sundar, SVP, and Chief Technology Officer – Key Accounts – Maveric Systems Limited, discusses how generative artificial intelligence (GenAI) is enhancing productivity and customer service in India's banking sector, while also addressing challenges in data management and cloud adoption. He emphasizes the role of edge computing and Maveric's focus on tech innovations to drive growth. Edited Excerpts:  

How is GenAI impacting banking experiences in India, especially considering your focus on customer experience? 

In India, there are two main segments where GenAI is being applied. First, productivity improvements, where companies are using AI internally. Enterprise use has become common, especially with robotic process automation (RPA) now enhanced by adding intelligence. Instead of being just rule-based, it's now more natural and generative, which simplifies and improves automation efficiency. 

On the customer experience side, we haven't seen much use yet. Most of the focus has been on agent assist, helping customer service agents respond to queries. Beyond that, not much AI has been applied, except in onboarding solutions, though India already has strong digital onboarding systems. 

For customer queries, particularly in the mortgage sector, AI is used to help with tasks like checking loan amounts, interest rates, and deregistration from mortgage documents. This process, which was manual and time-consuming, is now faster with AI tools that use free text and document search, providing more explainability. However, this is still being applied in limited areas, not on a large scale yet. 

As a service provider, where do you see opportunities for GenAI to evolve in banks and financial institutions beyond customer engagement? 

Some of the solutions we're building are based on government-driven initiatives (GNI), such as fraud analytics. For example, open banking, Unified Payments Interface (UPI), and peer-to-peer payments offered by National Payments Corporation of India (NPCI) in India and the Monetary Authority of Singapore are thriving in these markets, while other regions are just starting to adopt them.

As they open up, scams and spoofing are increasing, particularly in the US and Europe. In Europe, banks are now held liable for fraud under the Payment Services Directive 3 (PSD3) regulations. Additionally, even if a bank’s core system is down, transactions under 30 euros or pounds must still be processed. 

New regulations are driving the need for more advanced fraud analytics, and GNI is playing a key role. We are developing fraud analytics solutions using products like Nice and Actimize, allowing operations teams to detect anomalies in real time. Once detected, they can initiate two-factor authentication by holding the payment. We expect to see widespread use of this across different regions. 

The second focus is on anti-terrorism funding, anti-money laundering (AML), and Know Your Customer (KYC) regulations, which are becoming stricter in places like the US and the Middle East. While the UK and Europe have always had strong regulations, we're seeing new rules emerge around data lineage and KYC in these regions as well. We are partnering with products that help establish data lineage and ensure accurate reporting. This will help increase the sample size for periodic KYC validation from 10% to nearly 100%, improving operational efficiency. 
 
What challenges do banks face when using AI, and how does your company address these?

The main concern is data. None of the banks are willing to move their data or allow it to train public Large Language Models (LLMs). They want pre-trained LLMs that can be deployed on their premises or in a secure environment. 

The goal is to source open market data, train the LLM, and then deploy these models at the customer's location. 

Data is the primary issue for most banks. The second challenge is cloud adoption. While banks use the cloud for analytics and marketing data, they keep operational and transactional data on-premises. 

These factors are slowing down cloud adoption and scaling, while banks also face challenges with reducing bias and ensuring explainability with broader data sets.

Do you think there is less data available in India compared to the West for training LLMs? Are banks not using their data for this purpose? 

Correct, they need to start using their own data because the trained data isn't available. Most LLMs are trained with a focus on US-based data, leading to biases. The data used shapes the outputs, and Indian transaction patterns differ significantly from Western ones, especially in areas like spending and saving habits. This gap in data makes the limitations of models trained with non-Indian characteristics more noticeable. 

Many Chief Information Officers (CIOs) claim to be working on GNIs, but often, they are not fully applying their own data. Instead, they rely on ready-made data to show progress. As a result, using their own data is not as widespread as it should be. Additionally, there is no consolidated public data that reflects Indian characteristics, which could help service or technology providers offer solutions tailored to Indian customers.

How do you handle the challenges of migrating legacy systems to the cloud in regulated industries like banking? 

Banks outside India have taken a different approach to cloud adoption compared to Indian banks, which operate in a unique environment. 

Many countries have strict rules about keeping data within their borders. However, India is not as strict, though there is some discussion about it. In countries with strict data regulations, the banks we work with keep their core systems, like the system of records, on-premises. But they have moved layers like integration, customer experience, and front-end systems to the cloud. 

To protect sensitive data, especially customer PII (Personally Identifiable Information), banks are either tokenising or encrypting this data. They store unencrypted PII only on their on-premises systems, ensuring that even if there’s a breach, the damage is limited. This means data migration to the cloud hasn’t kept pace with the migration of applications or integration layers.

While banks have embraced cloud adoption, they're realising that costs have not been controlled. Horizontal scaling has increased their spending, especially in development and test environments, as costs are based on uptime, CPU usage, and storage. This has led to discussions, not just within banks but across enterprises, about whether they should reverse their cloud migration. It's a mix of success stories and reconsideration. 

This is where FinOps comes in. Many banks and enterprises have adopted FinOps teams to improve operational efficiency in cloud usage. 

In India, cloud adoption is still hybrid. Most banks remain on-premises, except for some cloud usage in application programming interfaces (APIs) like UPI. Unlike fintechs and startups, which are fully cloud-native, banks have legacy systems that make cloud migration costly without re-architecting their systems. Simply lifting and shifting to the cloud without containerisation won’t work. Fintechs and startups, being cloud-native, have an advantage in managing their environments more efficiently. 

What role do you see for edge computing in your future customer-facing solutions, considering its importance for real-time analytics in financial services? 

Edge computing is essential in cases like fraud analytics, where decentralised processing is needed. Traditionally, 40-50% of banks' activities used batch processing, but now they are shifting to real-time processing. Edge computing helps improve efficiency by making processes faster and more streamlined.

At the same time, regulations like GDPR in Europe and the California Consumer Privacy Act (CCPA) in the US impose stricter data security and privacy requirements, adding complexity to processing. Edge computing is widely used, both in the cloud and on-premises, to meet these demands. 

Your company aims to triple its revenue by 2025. What technological innovations or frameworks are you focusing on to ensure your platforms scale effectively? 

We are seeing four major trends in the market. First, banks are investing heavily in GenAI to improve both internal efficiency and customer experience. Second, many banks are focusing on modernising their second and third-generation systems. Over the past 20-30 years, they have built numerous applications with overlapping functionalities, some of which are underutilised. To streamline this, banks are rationalising and consolidating applications, and moving customisations out of their core systems to a microservices architecture. This makes upgrades easier and reduces complexity. 

Another significant trend is the increased spending on regulatory compliance, driven by stricter regulations. This includes investments in cybersecurity, data governance, and KYC validation processes. Lastly, banks have spent considerable resources over the past few years to enhance customer experience. Now, they are shifting their focus to monetising these improvements by using marketing technologies alongside their existing customer experience platforms. 

In response to these trends, we are transitioning from a competency-led approach to a service-line-led approach. Our focus is on simplifying operations, reducing technical debt, building lean core systems, and digitising and automating both functional and technical processes. We are also enabling third-party ecosystem integration and developing tools to accelerate these processes. In the application management services (AMS) space, we are applying cloud-based methodologies and building frameworks around site reliability and chaos engineering, moving from a reactive to a proactive approach. This is part of our effort to re-engineer AMS. 

On the regulatory side, we are partnering with products used for regulatory reporting to establish seamless data lineage and real-time AML and KYC solutions. Our strategy is to approach banks with service-led solutions rather than a competency-led focus, which we believe will help us triple our revenue by expanding services to existing customers. Additionally, we are aiming for 20-25% of our revenue to come from acquisitions in the US, UK/Europe, and India. By combining organic growth through service expansion and inorganic growth through acquisitions, we aim to drive substantial business growth. 

Do you see quantum computing or blockchain as having the potential to disrupt enterprise banking operations? How are you preparing to integrate these technologies into future solutions? 

Blockchain has been around, but it hasn't gained the same attention as GenAI. However, we are focusing on areas we discussed earlier, like data technologies and digital engineering, especially in marketing tech, where we're improving efficiency by applying digital engineering and enhancing analytics. 

We're investing heavily in this. Product management and domain expertise are key, especially in the regions we operate: Europe, the Middle East, Asia, and North America. We are building knowledge of regulations, customer behaviour, and product configurations specific to each region. 

Traditional digital skills, like cloud, automation, and quality engineering, are still important.