What Prasad demonstrated

Built a leakage-aware turbine yield regressor end to end: framed TEY as the target, excluded downstream CO/NOx emissions, trained a LinearRegression baseline and a RandomForest model, and validated on a genuine time-based holdout. The final model was evaluated honestly with held-out MAE, RMSE, and R2, and the accompanying model card documented intended use, limitations, and monitoring considerations.

• Leakage-Free Feature Selection
• Time-Based Holdout Evaluation
• Baseline-vs-Ensemble Comparison
• Plain-Language Model Communication