Sardine launches AI research lab, $375,000 in fellowships for fraud prevention

Sardine said it has launched Sardine AI Labs, an applied research group focused on frontier intelligence for fighting fraud and financial crime. The lab will work on production-grade AI models for fraud and financial crime that meet real-world latency, governance and explainability requirements. Sardine said early results from its foundational AI model for card transaction fraud showed a 68% improvement in fraud detection accuracy for a consumer card issuer and a 41% improvement for a business card issuer. The company said it is opening applications for up to 5 research fellowships for independent researchers to build on the lab's early work.

Sardine said it has set aside $375,000 in research fellowships. Fraud and financial crime prevention is a major use case where many academic researchers still lack the large-scale datasets needed to make progress, Sardine said. Sardine AI Labs is aimed at closing that gap by working on complex fraud and financial crime problems. The company said payment activity and financial-access behavior have distinct patterns, and that transaction histories, devices and IP addresses can be combined into behavioral profiles that foundation models can use to spot anomalies and potential fraud.

8 trillion in payments. Ranjan said Sardine AI Labs will focus on production-grade models that meet real-world latency, governance and explainability requirements. The company said financial crime prevention remains costly for financial institutions and adds to payment-processing expenses across card. Foundation models may help address persistent costs in ACH and wire transactions.

One example is sanctions screening, where consumers with common names or identities similar to those on sanctions lists can be incorrectly flagged, delaying onboarding for weeks while financial institutions carry out manual reviews. Determining whether two records refer to the same individual remains a complex problem. Another challenge is spotting money laundering, terrorist financing and transactions linked to drug trafficking among the billions of payments processed by banks.

Current systems can create false positives that delay legitimate transactions while still missing illicit activity, exposing financial institutions to significant regulatory and financial consequences. Sardine AI Labs will study AI research problems including modeling very large sequences of transaction and user-behavior events, transfer learning, adversarial robustness and explainability. The aim is to turn advances in AI into measurable protection against financial crime. The final sentence in the source is truncated: 'Grounded in Sardine’s'.

Sardine AI Labs said its new research lab is building models using device, identity, behavioral and transaction data to understand complex financial activity and spot attacks the models were not explicitly trained to detect. The lab also released early results from a foundation transformer model trained on complete cardholder transaction histories. Sardine said the card model was trained on about a billion transactions over the last 2 years from over a dozen card issuers.

The company said one of the biggest fraud-prevention challenges for new card issuers is the cold-start problem, because they initially lack enough training data to detect fraud reliably even though fraud rings often target newly released financial products first. In testing with issuers that were fully excluded from training, the model improved fraud detection accuracy by 68% for a consumer card issuer and 41% for a business card issuer versus standard machine-learning approaches. Sardine said the accuracy gains were consistent across 2.

Niranjan Shetty, head of data science at Sardine, said the key takeaway was that foundation models can learn directly from a user’s transaction behaviour, enabling more accurate fraud detection than methods that reduce that behaviour to tabular features. Shetty said those patterns transfer across financial institutions, so the model is not confined to one card programme, and said the next step is to make the intelligence fast, explainable and reliable enough for real-world risk decisions.

Sardine CEO and co-founder Soups Ranjan said frontier AI labs have shown the benefits of multimodal AI models that can draw intelligence from audio, video and text, and said the next breakthrough in fraud and financial-crime prevention would come from multimodal AI across device, identity, behavioural and transaction data. Ranjan said that is what excites him most about the potential of the company’s AI lab and that he looks forward to seeing what the research fellows build with the company. Sardine said the model architecture, training methodology and full results are set out in a new technical whitepaper.

The company also introduced the Sardine AI Fellowship Program, inviting independent researchers enrolled at universities in the US or Canada to extend the work through the new initiative. Sardine said the programme will select up to five fellows to study how models can generalize across card issuers, learn from multiple streams of financial activity, improve as data and model size scale, perform with limited customer history, improve identity matching, detect money-laundering patterns, support real-time inference in hundreds of milliseconds, and extend beyond fraud. Applications are open through December 13, 2026.

Shortlisted applicants will be notified in December, and the inaugural cohort of fellows will be announced in mid-January 2027. Sardine said its integrated fraud and financial crime solutions unify data across risk teams and support real-time fraud detection and automated compliance operations. It said more than 500 Global enterprises use Sardine to secure and grow their products, including category leaders in banking, fintech, wealth and retirement, HR and payroll, payments and software. Customers listed include FIS, Experian, National Bank of Canada, Nubank, GoDaddy, Deel, Gusto, Paylocity, Xero and ZoomInfo.

ai.