https://quantum-ui.com/blog Engineering blog from the Quantum team on AutoML, MLOps, feature stores and enterprise ML. en-us Fri, 24 Apr 2026 20:41:49 GMT https://quantum-ui.com/blog/poc-to-production-graveyard https://quantum-ui.com/blog/poc-to-production-graveyard Fri, 24 Apr 2026 00:00:00 GMT Your notebook model hit 0.94 AUC. The business loved the demo. Six months later, nothing shipped. Here is the map of where ML projects actually get killed, and what separates the survivors. noreply@quantum-ui.com (Bryam Camilo Acevedo) mlops production automl feature store https://quantum-ui.com/blog/data-leakage-silent-killer https://quantum-ui.com/blog/data-leakage-silent-killer Thu, 23 Apr 2026 00:00:00 GMT Your model hit 0.97 AUC on the test set. Then production happened and it dropped to 0.68. The gap between those numbers almost always has the same name. Here is how to find it before your launch day does. noreply@quantum-ui.com (Bryam Camilo Acevedo) data leakage model evaluation automl best practices https://quantum-ui.com/blog/xgboost-beats-neural-network https://quantum-ui.com/blog/xgboost-beats-neural-network Wed, 22 Apr 2026 00:00:00 GMT Tabular data is 80% of enterprise ML. Gradient-boosted trees are still the champion of tabular, and a well-tuned XGBoost will out-predict a transformer on most of your real problems. Here is the benchmark, and here is when to use each. noreply@quantum-ui.com (Bryam Camilo Acevedo) xgboost neural networks tabular benchmarks automl