Conducted an end-to-end statistical analysis of 1,000 NYC DOHMH restaurant inspection records covering 212 inspections and 195 establishments from 2016–2025. Built a Python pipeline for data cleaning, inspection-level aggregation, exploratory analysis, assumption checking, hypothesis testing, and visualization. Applied Shapiro-Wilk, Levene, Kruskal-Wallis, Mann-Whitney U, chi-square, and Spearman tests with effect sizes. Produced publication-quality visualizations and documented data-quality issues, methodological decisions, findings, and limitations. The project demonstrates practical skills in Python, Pandas, SciPy, statistical inference, data validation, and analytical reporting.

Developed a Python-based HR record-matching tool for a TechAbout assessment to reconcile applicant and employee records during system migration. The solution normalizes CNIC, email, phone numbers, and names, then applies a priority-based matching process: exact CNIC, exact email, exact phone, and fuzzy name similarity as a fallback. Each match receives a confidence level and explanation, while ambiguous cases are flagged for manual review instead of being silently assigned. The project includes audit-ready outputs, conflict handling, automated tests, and a modular Python structure using RapidFuzz.

Completed a practical machine-learning project portfolio during the Elevvo Pathways internship, covering four applied tasks in Python. Projects included customer segmentation using K-Means clustering with Elbow Method analysis, forest cover classification using Decision Trees and Logistic Regression with model evaluation and tuning, loan approval prediction using Logistic Regression and Decision Trees with SMOTE for imbalanced data, and Walmart sales forecasting using feature engineering, Linear Regression, XGBoost, and seasonal decomposition. The work demonstrates practical experience in preprocessing, modeling, evaluation, visualization, and applied machine learning using Python, Pandas, Scikit-learn, XGBoost, Statsmodels, and Jupyter/Google Colab.