Generative Model Engineering for Trading

This unit develops practical skills to engineer, train, and evaluate generative and hybrid AI models for trading and asset management. Students will convert legacy code, build data and feature pipelines, apply advanced generative model architectures and training techniques (GANs, WGANs, VAEs, flows), and assess model performance including quantization, sentiment analysis, and options-term-structure strategies in preparation for deployment.

Learning Objectives

• Apply techniques to translate MATLAB model code to reproducible Python implementations and validate equivalence on financial datasets
• Design and implement data engineering and feature-engineering pipelines for time-series, options, and sentiment datasets to prepare inputs for generative and discriminative models
• Train and optimize generative and hybrid architectures (GANs, WGANs, VAEs, coupling flows) using improved training techniques and model quantization to enhance stability and inference efficiency
• Evaluate experimental results, including model metrics, sentiment analysis outputs, and options term-structure arbitrage tests, and produce deployment-ready recommendations for asset management use cases