Machine Learning Hands-On

Subhra Das

Kolkata, India *************@*****.*** +91-808******* LinkedIN

Summary

Passionate Statistics and Computing postgraduate, skilled in data pre-processing, model evaluation, and exploratory analysis. Eager to contribute strong analytical thinking and hands-on project experience to real-world research and data-driven innovations. Education

Banaras Hindu University, M.Sc. in Statistics and Computing July 2025

• GPA: 6.67

Midnapore College, B.Sc. in Statistics July 2023

• GPA: 7.45

West Bengal Council of Higher Secondary Education, Intermediate Mar 2019

• 83.2 %

Experience

Research Intern, IIT BHU, Varanasi May 2024 – July 2024

• Developed a high-accuracy digit recognition system using Machine Learning on the MNIST dataset.

• Achieved a model accuracy of 99.2% by optimizing architecture and hyperparameters.

• Engineered a robust image preprocessing pipeline, enhancing model performance and generalization.

• Outcome: Created a foundational model for OCR applications, demonstrating deep learning proficiency. Projects

Social Media Sentiment & Engagement Intelligence Python, ML, NLP Apr 2025

• Analyzed 21,000+ tweets using NLP and ML models (Random Forest, SVM, Logistic Regression) to classify public sentiment with 82% accuracy, uncovering key perception trends.

• Forecasted engagement metrics using Facebook’s Prophet model, achieving a low MAE of 0.07, enabling data-driven content scheduling.

• Identified 313 anomalies in user engagement via Isolation Forest, providing early signals for proactive content strategy adjustments.

Credit Card Customer Churn Prediction Python, Imbalanced Learning Dec 2024

• Developed an end-to-end ML pipeline to predict customer attrition, achieving 96.6% accuracy, 0.90 precision, and a 0.991 ROC-AUC score.

• Engineered features and handled class imbalance to improve recall; used SHAP for model interpretability to identify top churn drivers.

• Enabled proactive retention campaigns, estimating $ 171K in annual savings by reducing customer attrition and protecting revenue.

Retail Demand Forecasting & Inventory Optimization Python, Time Series Analysis Dec 2024

• Forecasted 3-month demand for 50 items across 10 stores using 5 years of historical data, engineering time-series features (lags, rolling windows).

• Achieved high forecast reliability with R = 0.92, MAE = 6.1, and RMSE = 7.97.

• Forecasted sales of 2 .56M items, enabling smarter inventory planning, reducing stockouts and overstock risk. Certifications

• Complete Python Programming for Beginners - 2024 - Udemy

• Artificial Intelligence In Consulting & Project Management - Udemy

• NPTEL - Introduction to Biostatistics

Skills

• Languages & Databases: Python, R, SQL (PostgreSQL, BigQuery)

• Machine Learning & DL: Scikit-learn, TensorFlow, PyTorch, XGBoost

• Data Stack: Pandas, NumPy, Matplotlib, Seaborn, Plotly, Git, Jupyter

• Key Competencies: Regression Analysis, Statistical Analysis, Hypothesis Testing, Data-Driven Insights, Time Series Forecasting, NLP, Bayesian

• Soft Skills: Analytical Problem-Solving, Collaborative Research, Effective Communication, Leadership

Contact this candidate