Professional Summary
I hold an MCA in Data Science from Dayananda Sagar University, Bengaluru (2026). I build end-to-end analytics systems: raw data → Python cleaning & EDA → cloud data warehouse → automated pipeline → Power BI / Streamlit dashboards → actionable business recommendations.
Core Competencies
My technical foundation spans Python, SQL, Pandas, NumPy, Matplotlib, Seaborn, Power BI, Streamlit, Excel, MySQL, BigQuery, BigQuery ML, Supabase (Postgres), dbt, SQLAlchemy, and GitHub Actions (CI/CD). I focus on exploratory data analysis, SQL analytics (CTEs, window functions), RFM segmentation, cohort analysis, customer lifetime value, survival analysis (Kaplan-Meier, Cox Proportional Hazards via lifelines), and rigorous statistical validation (Welch's t-test, Shapiro-Wilk, Levene's, Mann-Whitney U, Chi-Square).
Featured Work
Two flagship projects show my analytical approach, deliberately built on different stacks to demonstrate range rather than repeating the same recipe:
-
Customer Revenue Intelligence System — End-to-end system on 779,407 UK e-commerce transactions (UCI Online Retail II): RFM segmentation, cohort retention, CLV, and revenue analysis in Python/SQL, loaded into a Supabase Postgres warehouse via an automated GitHub Actions pipeline with a 17-check data-quality gate. Added Kaplan-Meier and Cox Proportional Hazards survival analysis to model time-to-churn (concordance 0.87), and rebuilt the statistical testing with Welch's t-test, Shapiro-Wilk, and Levene's — catching a real false positive in the original Q4-seasonality significance claim, reported honestly rather than hidden. Delivered through a 4-page Power BI dashboard and a 6-page interactive Streamlit reporting app. Finding: the top two RFM segments (32% of customers) drive 80.23% of revenue, and 79% of customers never return after their first purchase. Live dashboard · Live app · GitHub
-
BFSI Loan Risk Analytics — Processed 1,342,942 lending records (91 engineered features, 0 nulls), discovering Grade G loans default at 49.50% compared to Grade A at 5.97%. Built infrastructure using MySQL, BigQuery, and dbt (3-layer pipeline, 9 models, 44/44 passing tests), with a BigQuery ML logistic regression model (ROC AUC 0.669, time-based train/eval split, explicit leakage/multicollinearity feature exclusion) scoring every loan's default probability. Added a local-LLM risk-narrative layer (Ollama, zero-cost, zero data leaving the machine) — with SQL/Python computing the actual highest/lowest-risk segments and the LLM restricted to phrasing them, after an early version let the LLM pick the extremes itself and got 2 of 4 wrong. Delivered through a 5-page Power BI dashboard. Live dashboard · GitHub
Job Search Status
I'm actively seeking Data Analyst roles (primary target), and I'm also open to Business Analyst and BI Analyst positions, in the BFSI and e-commerce analytics domains. Based in Bengaluru, targeting Bengaluru-based startups and IT/consulting firms as well as remote roles across India.
