ML Curriculum Project Evaluator & Review Simulator

Project Architecture Workbench

Configure training pipelines and test for real-world curriculum pitfalls

Interactive

Project Name & Domain Portfolio capstone project topic & problem domain

Data Split Strategy Pre-split imputation/scaling will trigger data leakage flags.

Primary Evaluation Metric High accuracy on skewed datasets (e.g., 95% majority) masks failed classification.

Model Architecture & Hyperparameters

Production Docker & Inference API FastAPI / Flask endpoint containerized with dependency lockfile

Model Drift & Telemetry Monitoring Evidently / Prometheus metrics tracking population distribution shifts

Live Pipeline Inspection

Senior ML Reviewer Audit

Automated inspection based on rigorous bootcamp & industry rubrics

Needs Revision

Curriculum Rubric Competency Breakdown Benchmark: ≥ 80 for Production Pass

Data Pipeline Leakage Check

Feature scaling and imputation occurred before train/test splitting. Test set statistics contaminated the training distribution!

Evaluation Metric Validation

Accuracy is misleading for imbalanced churn data; switch to Precision-Recall AUC.

Reviewer Actionable Fix

Apply stratified K-fold cross validation and separate scaling before feature imputation.