The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) LLMs for Equities Feature Forecasting at Two Sigma with Ben Wellington - #736
91 snips
Jun 17, 2025 
In this enlightening discussion, Ben Wellington, Deputy Head of Feature Forecasting at Two Sigma, shares his expertise in AI-driven equity feature forecasting. He delves into the intricacies of identifying and quantifying measurable features to improve predictive accuracy. The use of satellite imagery for data points like vehicle counts unveils unique insights. Ben emphasizes the importance of strict data timestamping to avoid temporal leakage and discusses the transformative role of large language models in enhancing data analysis. He also offers a glimpse into the future of agentic AI in finance.
AI Snips
Chapters
Transcript
Episode notes
Feature Hunting from Observations
- Ben Wellington illustrates feature creation by observing a help wanted sign and hypothesizing about hiring data's predictive value.
- He emphasizes the challenge of possessing historical data and the importance of timestamping to avoid temporal leakage.
LLMs Accelerate Feature Creation
- LLMs reduce feature creation time drastically from months to minutes, enabling fast hypothesis testing.
- This speed transforms previously unfeasible ideas, like detecting nose touching, into practical explorations.
Power of Embeddings Explored
- Embeddings transform words from isolated tokens into semantically rich vectors, enhancing model generalization.
- This allows models to infer relations between similar concepts, improving prediction beyond simple word counts.