Andrew Stewart

Builder / creator.

Projects

Engineering

earnings-signal-pipeline

Checks whether the tone of earnings calls predicts stock moves.

Role
Builder
Timeframe
2025
Accuracy
86%
Category
Engineering

Problem

Signals from language are hard to trust unless they line up with actual price moves.

What I built

A model trained on earnings-call language, joined with stock data to see if tone actually predicts how the stock moves.

PythonFinBERTPostgreSQLPolygon.ioFlaskPlotly

Outcome

86% accuracy, packaged as a small, self-contained example of using language data in finance.