The laboratory component of ECE-381: Applied Machine Learning is designed to provide students with a hands-on, experiential understanding of core machine learning techniques by working directly with edge computing devices, state-of-theart deep learning models, and interactive AI frameworks. This lab series emphasizes both foundational concepts and practical skills, bridging the gap between theoretical knowledge and real-world implementation.

Throughout the course, students will work primarily with the NVIDIA Jetson Orin Nano development kit in both standard and headless configurations. The labs begin with essential setup procedures, including Docker-based execution environments and Jupyter Notebooks, before progressing to advanced tasks involving real-time computer vision, object detection, image regression, segmentation, keypoint estimation, visual transformers, generative models, and large language models.

Each lab module focuses on a specific task aligned with modern machine learning pipelines. For example, students will build and train image classifiers using ResNet-18, implement coordinate regression for facial landmark detection, and perform multimodal inference using YOLOv11n and NanoOWL. Later labs introduce students to cutting-edge architectures such as Visual Transformers (ViT), Stable Diffusion for text-to-image generation, and TinyLLaMA-based LLMs for interactive natural language understanding and generation.

The primary objectives of these labs are to:

• Expose students to end-to-end ML workflows on embedded hardware.
• Foster the ability to design, train, and evaluate deep learning models for real-time applications.
• Enable students to explore multimodal AI systems through practical tasks.
• Encourage experimentation and iterative improvement using visual feedback and parameter tuning.

By the end of the lab series, students will have developed a robust skill set in implementing machine learning algorithms across a wide spectrum of tasks, gaining confidence in working with both vision and language models. The structured progression from classical classification to generative AI ensures that students acquire both breadth and depth in applied machine learning.