Projects
Work That Speaks
End-to-end data projects — from raw SQL to interactive dashboards and ML pipelines.
Vendor Performance Analysis — Retail Inventory & Sales
- Developed and optimised a complex SQL ETL pipeline to build an aggregated summary table from multiple tables; improved query performance using CTEs, significantly reducing processing time for large datasets.
- Conducted EDA and hypothesis testing in Python to evaluate vendor profitability, pricing strategy effectiveness, and inventory turnover.
- Identified over-dependence on top 10 vendors (65.7% of purchases) and uncovered $2.71M in unsold inventory from low-performing vendors, recommending diversification and inventory optimisation.
- Built an interactive Power BI dashboard visualising vendor performance, profit margins, and bulk purchasing impact (72% cost reduction).
SQLPythonPower BIEDAHypothesis Testing
Hospitality Analytics — AtliQ Hotels
- EDA on 10,000+ hotel booking records across multiple properties
- Identified top 3 drivers of booking cancellations
- Insights estimated to reduce cancellation rates by 15–20%
- Delivered dynamic pricing & demand forecasting recommendations
PythonPandasNumPyMatplotlibEDA
Vendor Invoice Intelligence System
- Built a dual-model machine learning system for freight prediction and invoice risk detection in finance workflows.
- Achieved R² ~97% (MAE ~24) in regression and ~89% accuracy (F1 ~0.82) in classification using Random Forest.
- Implemented end-to-end pipeline (SQL feature engineering → model training → inference) and deployed interactive Streamlit app enabling real-time invoice evaluation and decision support.
- Performed EDA & statistical testing (t-tests) to identify key drivers of invoice anomalies
- Deployed a Streamlit dashboard for real-time predictions and explainable insights
PythonMachine LearningRandom ForestSQLStreamlitGridSearchCV
