Jiaqi Wang integrated NPU acceleration into the Mindspeed Framework within the alibaba/ROLL repository, focusing on both training and inference workflows. Using Python and YAML, Jiaqi introduced new configuration files and updated model configs to support NPU-specific arguments and optimization pathways, enabling hardware acceleration for deep learning workloads. The work emphasized configuration management and NPU optimization, laying the foundation for improved throughput and latency in future deployments. Although no customer-facing bugs were addressed during this period, the feature development prioritized stability and traceability, aligning the repository for broader rollout and supporting scalable, high-performance deep learning applications on specialized hardware.