Foundations of Generative AI Trading

This unit introduces practical foundations for building AI-driven trading and asset management workflows: acquiring and preprocessing market data, computing core performance metrics, and converting strategy specifications into executable backtests. Students will also be introduced to supervised and unsupervised learning, basic generative model concepts (including diffusion models), and applied LLM-based sentiment analysis to prepare for more advanced model development and deployment in subsequent units.

Learning Objectives

• Analyze historical market data by retrieving adjusted closing prices, computing daily returns, plotting risk-free-rate trends, and calculating performance metrics such as Sharpe ratios using Python or MATLAB
• Apply software tools and workflows (installing Python environments, using provided MATLAB routines) to translate a written trading strategy specification into reproducible backtest code and compute portfolio constructs such as the efficient frontier and tangency portfolio
• Demonstrate core machine learning methods by implementing supervised and unsupervised learning pipelines and exploring representation complexity and basic generative model concepts (including diffusion models) to generate or augment trading signals
• Evaluate natural-language signals by performing sentiment analysis of Fed press conference speeches with large language models and integrate those sentiment features into backtests to assess predictive value