Applied Deep Learning Integration and Deployment

This capstone unit integrates applied deep learning tools, frameworks, and hardware to solve real-world problems across vision, speech, forecasting, and decision systems. Students will synthesize acceleration, cloud deployment, and responsible-practice concepts to prepare models and workflows for industry or advanced study.

Learning Objectives

• Analyze and compare deep learning frameworks and toolchains (PyTorch, MXNet, CNTK, Deeplearning4j, Neural Designer) and their suitability for tasks such as image generation, voice synthesis, game agents, and time-series forecasting
• Apply GPU acceleration and performance optimization techniques (CUDA, mixed precision, batching, profiling) and cloud-based workflows to develop and deploy models that meet target performance and scalability requirements
• Demonstrate end-to-end model development and deployment by building, documenting, and validating solutions for domain problems (solar savings estimation, building design, catastrophe forecasting, demographic prediction, and data mining) using appropriate libraries and services
• Evaluate ethical, safety, privacy, and workplace considerations (bias in demographic prediction, data privacy, model safety, and team practices) and propose mitigation strategies and documentation for responsible AI deployment