Projects

Problem
Stakeholders needed a clear, reliable way to compare intraday volatility models and understand which approach performed best on high-frequency market data. Tables/scripts and static plots made side-by-side comparison and communication difficult.

Data
SPY ETF intraday price data (15-minute frequency) with VIX as an exogenous volatility indicator; engineered features including log returns, rolling volatility, EWMA, and lagged volatility terms.

Build
Built an interactive Streamlit dashboard to benchmark GARCH, LSTM, and hybrid GARCH–LSTM–VIX variants using walk-forward evaluation. Visualized performance with MAE/RMSE and forecast-vs-actual comparisons, with structured explanations for non-technical readers.

Outcome
Enabled one-glance model comparison across accuracy and stability, reduced manual analysis by consolidating metrics and visuals in one place, and improved decision support through transparent, reproducible results.

Problem
What decision was hard because reporting was slow or unclear?

Data
List sources: Excel/CSV, SQL tables, APIs, operational systems.

Build
KPI model + measures + refresh workflow + validation checks.

Outcome
Add a measurable result: time saved, fewer errors, faster turnaround.

Problem
Mismatch across systems caused delays and rework.

Data
Two or more sources requiring reconciliation.

Build
Rules + exception report + audit trail + documentation (SOP).

Outcome
Fewer discrepancies and improved data trust.