t-learner

Taking causal inference to the extreme!

The package is developed for treatment recommendation & pairwise treatment individual effect estimation (ITE/CATE/HTE) when multiple treatment/intervention options exist. The package is still under development.

A repo with functions for building various COMs and GCOMs quickly.

This project aims to implement meta machine learning algorithms for causal inference

Uplift modeling on email marketing data (Hillstrom, 64k customers): estimating the causal effect of a campaign with a T-Learner (Random Forest) to target “persuadables”.

Causal Inference pipeline using T-Learners with LightGBM to estimate heterogeneous treatment effects (Uplift) in promotional campaigns.

Production-grade causal uplift modeling on 14M rows, benchmarks S-Learner, T-Learner, and FT-Transformer challengers on the Criteo dataset, with Optuna tuning, MLflow tracking, FastAPI + Docker + Google Cloud Run serving, and a Streamlit dashboard.

Uplift modeling for employee retention — who to keep, not who will leave. A T-learner recovers heterogeneous treatment effects and beats flight-risk targeting (semi-synthetic demo)

Causal Inference Engine using T-Learners (XGBoost) to optimize marketing ROI. Features: 3.2x Lift over random targeting, Behavior-Based Segmentation (Persuadables vs. Sleeping Dogs), and fully dockerized FastAPI/Streamlit architecture.