Sachin Kanchan

Data Analyst with extensive experience delivering business impact through Python and SQL. Built predictive models for lead conversion and demand forecasting, and led supplier optimization projects resulting in significant cost savings. Skilled in end-to-end data analysis using Pandas, creating advanced visualizations with Seaborn & Tableau, and experience with modern data stack principles, including Cloud Data Warehousing using Snowflake.

• HackerRank SQL Advanced Certification
• Google Data Analytics Professional Certificate

Credit Card Fraud Detection Model - 07/2025 - 08/2025

• Built a comprehensive machine learning solution to detect fraudulent credit card transactions in a highly imbalanced dataset (0.172% fraud rate).
• Implemented and compared 24 different model configurations using six algorithms - XGBoost, Random Forest, Decision Tree among others.
• Applied 3 oversampling techniques - ROS, SMOTE, ADASYN - with GridSearchCV hyperparameter optimization using 5-fold stratified CV and AUPRC evaluation metric.
• Final XGBoost model with Random Oversampling achieved 85% fraud recall and 94% precision, successfully identifying 83 out of 98 fraudulent transactions on the test set.

AI-Powered Text Summarization using RAG, NLP, Generative AI - 04/2025 - 06/2025

• Developed a Retrieval-Augmented Generation (RAG) pipeline using LangChain and ChromaDB to accurately query and extract insights from a large corpus of unstructured legal documents.
• Engineered a complete workflow including document chunking, embedding, and vector storage to enable efficient semantic retrieval of relevant information.
• Evaluated performance using RAGAS framework, achieving a high faithfulness score of 0.95, demonstrating the model's ability to generate answers grounded in source context.

Customer Churn Prediction - 03/2025 - 04/2025

• Developed a machine learning solution to predict churn for high-value telecom customers, enabling targeted retention strategies to reduce revenue loss.
• Dual-model approach: a high-performance Random Forest with PCA for prediction (AUC 0.86) and an interpretable Logistic Regression to identify key churn drivers.
• Engineered new features to capture critical changes in customer spending and usage patterns, which proved to be significant predictors.
• Delivered actionable recommendations — like spending-drop alerts and tiered loyalty programs — derived from the interpretable model's findings to guide business strategy.

Customer Segmentation for Targeted Marketing - 02/2025 - 03/2025

• Developed a Segmentation solution for an e-commerce retailer to enable a shift from generic to personalized marketing strategies.
• Engineered an RFM (Recency, Frequency, Monetary) feature set from raw transactional data to quantify customer value and engagement levels.
• Applied K-Means clustering with Elbow and Silhouette analysis to identify 3 optimal customer segments: high-value, mid-value, and lapsed customers.
• Delivered a strategic framework for targeted marketing, enabling tailored loyalty programs for top-tier customers and specific win-back campaigns to reduce churn.

Lead Conversion Predictive Analytics - 01/2025 - 02/2025

• Developed a logistic regression model to identify 'Hot Leads' for an education company, addressing a low conversion rate and aiming for an 80% target.
• Engineered a robust feature set through extensive data cleaning, outlier treatment, and RFE to select the most impactful predictors for the model.
• Achieved about 80% accuracy and a ROC-AUC of 0.85 on the test set, demonstrating exceptional model performance and reliability.
• Designed and implemented a lead scoring system (0-100) based on model probabilities, enabling the sales team to prioritize efforts on high-potential leads.

• Data, Financial Markets, Photography