Data Analyst building SQL pipelines and BI dashboards

Decision Intelligence Experimentation Platform

• Engineered an end-to-end experimentation system using PostgreSQL, dbt, and Python, simulating 50K users, 280K+ events, and real-world behavioral patterns.

• Built a dbt-powered metrics layer transforming raw event data into experiment-ready datasets with conversion rate, lift, and AOV calculations.

• Implemented statistical hypothesis testing (z-test) to evaluate experiment significance and automate product decisions based on p-value and lift.

• Developed a segment-aware decision layer identifying performance differences across platform and geography, enabling data-driven rollout strategies.

• Delivered a Streamlit dashboard presenting experiment metrics, statistical outcomes, and business-ready insights in a unified interface.

AI Data Governance Model

• Built an AI-assisted data governance system using PostgreSQL, dbt, and Python to monitor datasets and enforce automated data quality rules on GitHub Archive data.

• Designed a governance metadata layer (dataset registry, rule catalog, rule runs, health scoring, incident tracking) to operationalize dataset monitoring.

• Implemented a rule execution engine detecting data quality failures and generating incidents with severity tracking.

• Developed an AI-based incident explanation assistant to translate data issues into clear, actionable insights for stakeholders.

Retail Analytics Dashboard

• Architected a data pipeline to process and transform 541K+ retail transactions into analysis-ready datasets using SQL and Python.

• Engineered SQL aggregations and indexing strategies reducing dashboard load times by 40%.

• Implemented ABC inventory analysis identifying the top 20% of products driving 75% of revenue, enabling data-driven stock decisions.

• Built an automated Streamlit and Plotly dashboard replacing manual reporting workflows, saving 3–4 hours/week.

Customer Churn Prediction

• Developed an end-to-end churn prediction pipeline using XGBoost on 7,043 customers and 21 features achieving ROC-AUC 0.84.

• Engineered predictive features and applied resampling techniques to address class imbalance improving recall by 18%.

• Applied SHAP explainability to identify key churn drivers and translate model outputs into actionable retention strategies.