Show HN: Algeon – Compose GPU graph, vector, and ML algorithms in SQL (github.com)

Algeon is a groundbreaking GPU-accelerated query engine that integrates advanced algorithms from NVIDIA's suite—including cuGraph, cuVS, and cuML—into SQL, allowing for efficient execution of graph, vector, and machine learning algorithms directly within SQL queries. Built on Apache DataFusion, Algeon enables users to execute complex data manipulations and analyses seamlessly in a single query, enhancing the capabilities of analysts who require ad-hoc investigations without the need for cumbersome export scripts. This innovative approach supports the exploration of data relationships and patterns that would typically go unnoticed, promoting iterative testing and hypothesis generation. The significance of Algeon lies in its ability to streamline the analytic process, particularly for tasks involving large datasets on lakehouse architectures, by utilizing GPU resources for computationally intensive operations. Analysts can directly engage with both data input and algorithm execution via SQL statements, increasing efficiency and accessibility. Although Algeon currently has limitations—such as a maximum working set size constrained by GPU memory and compatibility primarily with Parquet data—it paves the way for more integrated and dynamic data analysis environments that blend SQL and machine learning. Algeon is still experimental, and as it evolves, it promises to enhance how data scientists and AI agents interact with complex datasets.