Your LLM Is Still Thinking. Java Already Detected the Fraud

Java powers much of the financial world. Now it's time to put AI to work on the JVM.

Fraud detection, risk analysis, and real-time financial decisions demand more than accurate models. They require high throughput, low latency, predictable performance, and enterprise-grade security. And they need to work with the Java applications and infrastructure already running the business.

In this session, we'll explore how modern Java delivers high-performance AI without leaving the JVM. We'll cover GPU acceleration with the Foreign Function & Memory API, CPU optimization with the Vector API, and efficient model inference for workloads where LLMs are too slow, too expensive, or the wrong tool.

Using a real-world fraud detection scenario, we'll demonstrate how Java developers can build and deploy fast, scalable AI solutions using familiar tools, frameworks, and infrastructure.

**Suite 1418**

6:00-6:30pm Pizza/refreshments

6:30-8:00pm Using Java for Fraud Detection without an LLM

**Zoran Ševarac** is a Java Champion, AI and machine learning expert, and CEO of Deep Netts, where he focuses on making AI faster and more accessible to Java developers.

He is a Full Professor at the University of Belgrade, leading its Laboratory for Artificial Intelligence. With over 20 years of experience, Zoran created the award-winning Neuroph neural network framework, co-authored JSR 381 (Visual Recognition API), and serves on the Java Community Process Executive Committee and its AI Working Group. A longtime Java community contributor, he frequently speaks at JavaOne, Devoxx, QCon, and other major developer conferences.