Today’s Patent – Rapid Predictive Analysis with Distributed Computational Graphs

The said invention (US12137123B1) was invented by Jason Crabtree, Andrew Sellers, and patented by USPTO on July 21^th, 2024. Currently, it stands assigned to the Qomplx Inc.

This invention relates to a system for predictive analysis of large-scale streaming and batch data using a distributed computational graph. The system includes a cluster of computer systems divided into two distinct groups.

One group processes incoming data in real time through configurable transformation pipelines that generate intermediate outputs. These outputs are passed along the graph and further processed by other systems in the cluster. Stored data is analyzed for patterns and trends, while streaming data is transformed on the fly. A controller component monitors performance, dynamically reconfigures pipelines as needed, and ensures system responsiveness.

Each node in the system stores a portion of the computational graph and executes transformation logic specific to its assigned role. As data flows through the system, monitoring software—assisted by machine learning algorithms—detects when new pipelines are needed and automatically assigns processing resources from the second group to handle the additional load. The architecture supports both real-time and batch processing, enabling scalable, adaptive analytics on massive data sets with minimal manual intervention.