A reproducible ML study of Codeforces difficulty prediction, solved-count exposure effects, cold-start limits, and temporal validation.

Temporal partitioning and backtesting for time-correlated datasets

R Shiny dashboard demonstrating validation-first analytics for clinical trial duration forecasting. Random split R² = 0.84 vs time-based R² = 0.04—why validation strategy matters more than model selection.

Production-aware credit-card fraud detection. Temporal splits, cost-sensitive thresholds, calibrated probabilities, Optuna tuning, and a Streamlit dashboard.

Stage359: Real Public Key Fingerprint Validation Layer with Stage358 temporal validity binding, SHA256 fingerprint metadata checks, key identity consistency, and future timestamp/revocation proof readiness.

Reproducible temporal-validation pipeline for suicidal-ideation prediction in employed U.S. adults across 2015-2023 NSDUH data.

Reproducible ML pipeline evaluating temporal leakage in Expected Pass Turnovers (xPT) models for football analytics. Compares 4 algorithms (mixed-effects logistic, penalised logistic, random forest, XGBoost) across leakage-inclusive and leakage-corrected feature sets. Supporting code for manuscript under review.

Stage358: Signing-Time Validity Enforcement Layer with key lifecycle temporal verification, Stage357 receipt binding, validity-window checks, revocation-time checks, and fail-closed temporal decisions.

Predicting lower-cost carrier entry into U.S. airline routes with temporal validation, XGBoost, and route-level market-intelligence scoring.

Five-day volatility forecast for the S&P 500 using XGBoost with strict time validation. Full pipeline: data cleaning, 29 technical indicators, 5-fold TimeSeriesSplit, 5 independent XGBoost regressors, and results visualization.

Production-style PySpark lakehouse for transaction risk modeling, fraud detection, and temporal model validation.

machine learning-based crop yield prediction with advanced feature engineering, temporal validation (1997–2020), model comparison, ablation study, and interactive Streamlit deployment.

A synthetic physical inconsistency detector using DINOv2 residual probes, multi-view validation, temporal sync drift checks, semantic mismatch detection, and counterfactual intervention generation.

Drift-aware training window extension assessment for production ML pipelines. Claude Code skill.

Methodology and aggregate results for leakage-safe six-month groundwater forecasting across 37 Bengaluru wards

Temporal cross-validation for insurance pricing - respects policy time structure, CatBoost, Polars