Ep. 365: Stefan Jansen on Agentic AI, ML Workflows, and the Evolution of Machine Learning for Trading

Applying machine learning and AI to financial trading presents unique challenges compared to general business applications, primarily due to low signal-to-noise ratios, limited historical data, and the necessity for rigorous, non-stationary time series analysis. Stefan Jansen, founder of Applied AI, emphasizes that successful trading strategies require a disciplined, end-to-end workflow—from hypothesis generation and feature engineering to robust backtesting and production monitoring. While large language models and autonomous agents offer significant potential for automating research, coding, and data processing, they do not replace human judgment in defining goals or managing complex, long-term projects. Instead, the future of quantitative finance lies in human-led, agent-assisted workflows that prioritize explainability, auditability, and the integration of domain-specific knowledge to navigate the inherent risks of competitive market environments.