← All Case Studies
2,600+ Stocks
Up From 3-4 Manual Reviews Per Week
ML-Driven Value Investing Model
Machine LearningFinancial ServicesPythonAutomation
Context
Private fund manager (~$20M AUM) could only evaluate 3-4 stocks weekly using manual 13F analysis. Needed to scale research capacity without adding headcount.
Challenge
The manual process of analyzing institutional investor filings was too slow and limited the fund's ability to identify opportunities across a broad universe of stocks.
Approach
- 01Automated 13F ingestion via APIs to flag stocks where institutional positions shifted 25%+ up or down
- 02Layered fundamental screens on flagged stocks to filter for quality
- 03Built a model across ~2,600 stocks (NASDAQ, NYSE, Russell 2000) using 20 years of quarterly fundamental and macroeconomic data
- 04Applied CatBoost classifier after feature engineering on fundamental and econometric ratios
- 05Built comprehensive reporting tool in Python pulling financials, running valuations, and using LLMs to summarize and generate buy/hold/pass recommendations
Results
Research capacity scaled from 3-4 stocks per week to systematic coverage of 2,600+
13F ingestion fully automated, flagging institutional position shifts of 25%+ without manual review
Freed the client from waiting for 13F disclosures
Key Insight
"The real value wasn't just the model - it was freeing them from waiting for 13F disclosures and giving them a systematic, repeatable process."