MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641106534 A) filed by S S Imambi; Santosh Kumar; and Koneru Lakshmaiah Education Foundation on September 04, 2026, for Optimization-Aware Distributed Gpu Orchestration System For Heterogeneous Clinical Data Processing.
Inventors include S S Imambi; Santosh Kumar; and Koneru Lakshmaiah Education Foundation.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: Optimization-Aware Distributed GPU Orchestration System for Heterogeneous Clinical Data Processing. A computing system processes heterogeneous data on a cluster containing CPUs, GPUs, host and device memory, and a network interconnect. A data orchestrator forms synchronized, complexity-balanced chunks. A runtime optimization controller evaluates feasible execution configurations using workload descriptors and live telemetry, filters dominated configurations and selects a configuration defining a feature subset, chunk size, device placement, parallelism mode, precision, memory policy and fusion strategy. A GPU planner overlaps transfers and kernels, while a scheduler and memory manager implement the selected configuration. A monitoring engine compares predicted and observed performance and, subject to stability guards, feeds the difference back to update the cost model and reconfigure subsequent chunks. The closed loop improves utilization of computing and memory resources and reduces latency and communication overhead. An aligned EHR, ECG and chest-radiograph clinical pipeline is an illustrative deployment.
Disclaimer: Curated by HT Syndication.
