Does Model Diversity Actually Matter in an ML Ensemble?
Why combining different models doesn't guarantee a better ensemble: correlated errors, complementarity, underfitting, and validation on the full system.
Ibexcode
Publications
Our Articles
Technical articles, lessons learned, architecture decisions and insights from real projects.
Machine Learning and Human Learning: Surprisingly Similar Problems
Memorization, generalization, evaluation, errors, and redundant knowledge: what certain Machine Learning problems reveal about learning itself.
Ibexcode
From Notebook to Production: Building a Robust ML Pipeline with Airflow and CI/CD
From notebook to production: orchestrating an ML pipeline with Apache Airflow, data validation, CI/CD, deployment, and observability.
Ibexcode
When an ML Improvement Doesn't Survive Validation
Why an apparent ML gain can disappear under more rigorous validation: statistical uncertainty, selection bias, redundancy, and system-level impact.
Ibexcode
Building a Temporal ML Ranking System Without Data Leakage
How to avoid data leakage in a temporal ML ranking system: future-only validation, OOF predictions, stacking, and calibration.
Ibexcode
Building an End-to-End ML Product as a Solo Developer: From Data to Mobile App
A firsthand account of building a complete ML product: data engineering, ranking, MLOps, cloud backend, CI/CD, and a Flutter app.
Ibexcode