Bridging reliability and high performance in AI execution with radiation-hardened co-processing

Radiation-hardened hardware is essential for mission-critical domains such as space, where high dependability is required. However, this requirement significantly limits the range of available components suitable for modern AI-intensive workloads—especially those involving real-time AI-based classification. To overcome this limitation, we propose using the High Performance Data Processor (HPDP) as a radiation-hardened, low-power co-processor combined with an optimized AI framework for efficient data processing.The HPDP features dynamic reconfiguration and a dataflow-oriented architecture, making it well-suited for AI applications that require low-latency and high-throughput streaming data processing. To leverage the capabilities of the HPDP as a coprocessor, we integrated it with Klepsydra’s AI-runtime inference framework, which, due to its lock-free execution and efficient resource management, significantly enhances data processing throughput without increasing power consumption.Our approach involves configuring the HPDP as a dedicated mathematical backend, allowing the AI framework to orchestrate and run workloads directly on this co-processor. This paper presents the results from our implementation, including the target application domain, AI pipeline, architectural highlights of the HPDP, and performance analysis. The proposed solution represents a significant advancement in executing AI on radiation-hardened platforms, positioning the HPDP as a versatile and reprogrammable coprocessor suitable for AI execution Harsh environments.