In March 2026, Siyuan Peng developed a per-session thread pool work callbacks API for the microsoft/onnxruntime repository, focusing on enhancing telemetry and resource management for per-session workloads. Leveraging C and C++ with expertise in API design and concurrency management, Siyuan implemented a policy-based thread pool architecture that introduces callback hooks for various thread pool lifecycle events while ensuring zero overhead when disabled. The solution included comprehensive wiring from environment to session, robust test coverage, and seamless integration with existing threading infrastructure. This work enables actionable diagnostics and performance instrumentation, addressing production observability needs with a well-architected, maintainable approach.