← All work

Applied ML · Data · 2025

Chemistry Outcome Modeling

A model built for Emory's Chemistry Unbound program to flag students who might struggle on the DUCK exam early, using placement scores, GPA features, course grades, and standardized-test data.

Exploratory modeling figure from the chemistry outcome-modeling project.
An exploratory figure from the modeling workflow.

Problem

Support programs need a signal early, well before the final exam results come in.

Approach

I pulled placement exams, GPA features, course grades, and a few model families into one prediction workflow.

Outcome

An early-warning workflow that turns past records into a heads-up for who to reach out to, trained on 223 students and scored across 2,866 downstream.