I analyze structured data to support operational and financial decisions.
My projects focus on forecasting demand, detecting risk patterns, and building dashboards that make performance trade-offs clear. I start with a business question and aim to end with insights that improve planning, reduce risk, or increase efficiency.
I value practical analysis, the goal is to reduce uncertainty and support better decisions.
Core Skills
Analytics
Forecasting
Classification
Modeling
Anomaly Detection
Tools
Power BI
Excel
Git
Languages
SQL
Python
Projects
Each project is built around a clear business question and ends with measurable outcomes.
Reduced false positives by 15% in a fraud detection model while maintaining high recall on imbalanced transaction data
Modeled short-term energy demand variability to support operational capacity planning
Analyzed 10,000+ app records to identify high-performing revenue and rating patterns
Built interactive dashboards to monitor energy usage trends and support efficiency-focused decisions
Approach
Start with the business question →
structure the data →
test assumptions →
deliver insights that inform decisions.
Energy Demand Forecasting
Business Question: How can energy demand variability affect short-term capacity and cost planning in data center operations?
Approach: Modeled baseline vs. renewable adoption scenarios using historical energy data. Structured datasets for comparative analysis and built interactive reporting views.
Key Insight: Scenario modeling indicated up to ~70% potential emissions reduction at 80% renewable penetration, with clear regional demand differences impacting planning assumptions.
Outcome: Delivered an interactive Power BI dashboard enabling comparison of demand patterns and supporting structured planning discussions.
Business Question: How can fraud risk be detected reliably in highly imbalanced transaction data without overwhelming investigation teams?
Dataset: 284,807 transactions (0.17% fraud rate)
Approach: Trained and compared multiple classification models (LogReg, Random Forest, SVM), applied SMOTE balancing, tuned hyperparameters, and evaluated using F1-score and ROC metrics.
Key Result: Random Forest achieved ~90.6% F1-score, balancing precision and recall to reduce false positives while maintaining fraud detection sensitivity.
Outcome: Produced a structured performance summary showing where financial risk controls could be tightened.