Data Processing Units (DPUs) enable computation to be offloaded closer to the network data path, reducing host CPU overhead and supporting responsive in-network services. This work presents a hands-on lab series developed by the Cyberinfrastructure Lab at the University of South Carolina for programming the Data Path Accelerator (DPA) on NVIDIA BlueField-3 DPUs. Using NVIDIA DOCA SDK and FlexIO, the labs introduce DPA programming incrementally, covering host-DPA synchronization, asynchronous kernel execution, memory transfers, multithreaded execution, software- and hardware-triggered wake-ups, completion queues, and live packet reception. Each lab focuses on one new concept and has been built and verified on physical BlueField-3 hardware. The series gives learners practical experience writing, building, and validating DPA applications that react directly to hardware events. These skills provide a foundation for advanced in-network services such as telemetry, packet processing, rate limiting, and firewalling.