Muskan Sharma

I'm building predictive models for operational and credit-risk decisions, from supply-chain delivery risk to real-time loan approval.

For a supply-chain delivery risk project, I analyzed 172K+ real orders using Python, pandas, scikit-learn, LightGBM, and SMOTE. My leakage-safe time-based model achieved 0.86 PR-AUC and 0.78 ROC-AUC, while identifying scheduled shipment duration as the leading risk driver through SHAP analysis.

I quantified $18.08 million in at-risk sales and designed a threshold where the top 20% of high-risk orders captured 34.3% of late deliveries at 98% precision. I also deployed a Flask loan approval application on Render after comparing seven classifiers and achieving 0.95 ROC-AUC.

Alongside my analytics work, I've gained exposure to steelmaking and continuous casting operations at SAIL, and I lead sponsorship, public relations, logistics, and student coordination as Vice President of the Society of Industrial and Management Engineering.