Quantitative Research using Snowflake, Machine Learning and Artificial Intelligence

Introduction, Set up, Workshop Overview

• Goal: Outline workshop objectives, set expectations, introduce data sources and the Snowflake Financial Data Cloud + AI Platform

• Topics:
  • Overview of Snowflake, capabilities, data sources (e.g., historical prices, earnings transcripts)
  • Brief intro to tools & libraries (Snowflake, SQL, Python, Snow Pandas, LLMs, Cortex AI, vector embeddings, vector search)
  • Agentic orchestration in quant workflows

Session 1 | From Factor Discovery to Portfolio Construction – End-to-End Quantitative ML on Snowflake

• Goal: Demonstrate how Snowflake’s unified AI/ML platform replaces multiple vendor tools for building, validating, and deploying a quantitative factor strategy — from raw market data to optimised portfolio.

• • Topics:
    • Constructing cross-sectional equity factors with Feature Store
    • Augmenting traditional factors with sentiment from earnings transcripts using Cortex AI Functions — and testing whether alternative data adds marginal alpha.
    • Validating factor premia with Fama-MacBeth regressions, thennon-linear factor interactions with XGBoost and SHAP explainability.
    • Model versioning, experiment tracking, and drift monitoring with Snowflake ML Registry and Model Monitor.
    • Accelerating Portfolio Optimization with NVIDIA Blackwell Computer

Coffee Break

Session 2 | Agentic Investment Data Analysis 

• Goal: Demonstrate how to deploy agents across the quantitative research lifecycle

• Topics:
  • Preparing the data
  • Semantic views and Vector Search
  • Deploying Agents across data sets
  • Real world results

• Hands-On Activity

Lunch

Session 3 | S&P – Lazy Prices (90 minutes)

• Goal: Show how to combine insights from structured and unstructured data.

• LQG Workshop: Not So Lazy Prices:
  • This workshop explores enhancing the original “Lazy Prices” strategy by adding an LLM layer to distinguish substantive risk changes from cosmetic textual updates in MD&A and Risk Factor disclosures.

• Agenda Points:
  • Lazy Prices Framework Review: Recap Cohen et al.’s finding that MD&A and Risk Factor textual changes predict negative returns
  • The LLM Enhancement Hypothesis: Explore how semantic analysis could filter non-incremental changes to improve short portfolio hit rates
  • Methodology Discussion: Approaches for distinguishing material risk disclosures from boilerplate updates using language models
  • Research Design & Next Steps: Framework for testing the strategy and measuring potential alpha improvements

• Hands-On Activity

Q&A, Wrap-Up, and Next Steps