The global financial industry is facing a significant challenge, with credit card transaction fraud losses projected to reach a staggering $43 billion by 2026. In response, NVIDIA has introduced a groundbreaking AI workflow designed specifically for fraud detection, operating on Amazon Web Services (AWS). This new solution aims to mitigate risks and save costs for financial services organizations through advanced data processing and algorithms.

Unveiled at the Money20/20 fintech conference, the NVIDIA AI workflow enables financial institutions to detect subtle patterns and anomalies in transaction data that reflect user behavior. By improving the accuracy of fraud detection, this workflow significantly reduces the occurrence of false positives compared to traditional methods.

Organizations can easily transition their fraud detection systems from conventional computing to accelerated computing by utilizing the NVIDIA AI Enterprise software platform alongside NVIDIA GPU instances. Businesses that adopt comprehensive machine learning strategies have seen an estimated 40% improvement in fraud detection accuracy. This enhancement allows them to identify and thwart fraud more swiftly, thereby reducing potential losses.

Prominent financial institutions such as American Express and Capital One are already leveraging AI to develop proprietary solutions aimed at mitigating fraud and enhancing customer security. The newly launched NVIDIA workflow not only accelerates data processing and model training but also provides an all-in-one software solution powered by NVIDIA AI.

Currently optimized for credit card transaction fraud, this workflow has the potential to adapt to various use cases, including new account fraud, account takeover, and money laundering. As AI models continue to evolve in complexity and scale, the demand for efficient computing power becomes increasingly critical for organizations in the financial sector.

Traditional data science pipelines often lack the necessary computational acceleration needed to process the vast amounts of data required for effective fraud detection. By utilizing the NVIDIA RAPIDS Accelerator for Apache Spark, payment companies can significantly reduce data processing times and lower costs.

To manage large datasets and deliver real-time AI performance with complex models, financial institutions are increasingly turning to NVIDIA’s AI and accelerated computing platforms. One significant innovation within the new workflow involves enhancing traditional gradient-boosted decision trees, commonly used in fraud detection, by incorporating graph neural network (GNN) embeddings. This addition helps to minimize false positives.

The GNN embeddings integrate with XGBoost, a popular machine learning algorithm, to create and train robust models. These models can be orchestrated using the NVIDIA Morpheus Runtime Core library and NVIDIA Triton Inference Server, which facilitate real-time inferencing. Morpheus ensures secure data inspection and classification, tagging incoming data with patterns and flagging suspicious activities. Meanwhile, Triton optimizes the deployment of various AI models, enhancing throughput and latency.

The adoption of AI in financial services is becoming increasingly critical, especially as many North American institutions report rising online fraud losses. American Express, which began using AI for fraud prevention in 2010, now monitors all customer transactions globally in real time, generating fraud alerts in mere milliseconds. By employing advanced algorithms, including those utilizing NVIDIA’s platform, American Express has significantly enhanced its model accuracy.

European digital bank Bunq is also innovating with AI, using generative AI and large language models to detect fraud and money laundering. Their AI-powered transaction monitoring has achieved nearly 100 times faster model training speeds with NVIDIA’s accelerated computing.

Recently, BNY became the first major bank to implement an NVIDIA DGX SuperPOD with DGX H100 systems, paving the way for solutions that enhance fraud detection and other applications. Now, systems integrators, software vendors, and cloud service providers can incorporate this new NVIDIA AI workflow into their financial services applications, enhancing security for customers' money, identities, and digital accounts.