Jashanpreet
Singh.

Predict failures across 22 turbines and estimate Remaining Useful Life from 1.2M+ SCADA records with severe class imbalance.

Dual model: 30-day failure classification + RUL regression. Engineered physics-based features (thermal deltas, drivetrain ratio, power-wind cubic efficiency, turbulence, rolling/lag windows). Cost-sensitive boosting with threshold optimization on a ~1:23 imbalance.

Optimized PR-AUC / Recall / F1; SHAP explainability per turbine; automated technician-scheduling agent in n8n.

Forecast next 3-day Nifty stock prices using 12 years of history with meaningful temporal signal.

30-day rolling window inputs into an LSTM with attention mechanism to weight temporal dependencies dynamically.

End-to-end pipeline from yfinance ingestion to trained model with evaluation on hold-out windows.

Consolidate fragmented CRM + ERP data into a queryable, governed analytics layer.

Bronze / Silver / Gold Medallion Architecture on Databricks + PySpark. ETL workflows with SQL CTEs for validation, standardization, and integration.

Reliable Gold layer supporting KPI tracking and reporting across business domains.