In November 2025, William Lee enhanced the google-ai-edge/LiteRT repository by enabling LogSoftMax and ScatterNd operations within the Qualcomm AI Engine, expanding supported tensor operations for edge AI workloads. He implemented these features using C++ and MLIR, focusing on robust integration and correctness through comprehensive MLIR and unit testing. William also refined the leaky ReLU path to improve engine stability and reliability. His work addressed model compatibility and runtime performance on Qualcomm hardware, with all major test suites passing and no customer-visible defects reported. This contribution improved deployment readiness and broadened model support for on-device machine learning inference scenarios.