MSc in Data Science & Analytics. Nine projects across marketing, retail, fintech, and applied ML, built to find the number that changes the decision.
I work end-to-end with data: cleaning and validating it, writing SQL and Python to analyze it, and building Tableau and Power BI dashboards that make it usable for decisions. My project work spans applied ML (churn classification, customer segmentation, anomaly detection), AI-augmented analytics (LLM-generated insights and recommendations, RAG-based chat interfaces), and business-facing analytics (marketing spend, retail pricing, fintech risk), alongside internship experience turning raw engineering and telemetry data into structured, reportable formats.
Currently looking at Data Analyst, Business Analyst, Data Science, and adjacent roles across India and the UAE, open to whichever team fits best.
200K+ AdTech records across 6 channels, tracing 109M+ clicks back to a 14.04% average CTR.
View on GitHub →Retail SQL analysis using window functions and CTEs, surfacing a counter-intuitive discount-effectiveness insight.
View on GitHub →9,977 transactions engineered into an explainable risk score, classified High-Risk vs. Not with a Random Forest model, and explained through Groq-generated recommendations and a RAG-lite Streamlit chat assistant.
View on GitHub →K-Means segmentation into 4 groups, XGBoost churn classification, and Ridge/Lasso credit-limit regression with hypothesis testing.
View on GitHub →3,051 weekly records across 26 divisions and 4 channels; Z-score flagging caught 267 anomalies and a Q4 spike to 131M vs. a ~47M average.
View on GitHub →A unified pipeline combining KNN regression, logistic regression, and K-Means (K=5), engineered into anomaly scores that surfaced high-risk and high-spend segments.
View on GitHub →A retrieval-augmented Q&A API over documents, served through FastAPI with ChromaDB and Groq.
View on GitHub →An upgraded assistant with LangChain 1.0, live web search, conversational memory, and dynamic file upload, with a Streamlit chat UI.
View on GitHub →ESC-50 sound classification comparing MFCC baselines against YAMNet transfer learning, deployed as a Flask app.
View on GitHub →