I build risk models, stress testing frameworks, and financial analytics systems that turn complex market data into actionable risk intelligence.
Six synthetic risk factor series with three injected data faults (frozen feed, bad print, gap). A gradient boosting detector trained on synthetic faults finds and names all three in a fraction of a second, a repair ladder is scored proposal by proposal against KS and VaR-impact guardrails, and a mask-and-recover harness benchmarks a random forest against linear methods on tail preservation. The finding: 99% VaR barely moves on corrupted data (+0.2%) while expected shortfall moves +71%, so the headline risk number is not a data alarm. Historical simulation VaR, stressed VaR window search, sensitivities, backtesting, and an LLM morning report held behind a number-check guardrail.
First line risk view of a synthetic agency securities lending book: daily client exposure monitor with margin calls, exposure limits and a non-standard collateral cap, a backtest of the flat 102/105 house haircut against a volatility-scaled haircut zoned with the Basel traffic light, and a counterparty fire drill that closes out a defaulted borrower over its worst historical two-day window with netting.
Second line of defense view of a bank's Banking book: balance sheet change attribution to LCR and NSFR, a limit framework with breach log, internal stress projection reconciled to the regulatory ratios by segment, and an evidence-based challenge of proposed stress assumptions with a benchmarked gradient boosting model as witness.
IDL monitoring dashboard with BCBS 248 alignment, ML net flow forecasting with GradientBoosting, six stress scenarios (counterparty delay, CCP margin spike, market stress), severity-classified playbook simulation, and channel analytics across Fedwire, CHIPS, ACH, Fed Securities, and CCP margin.
Liquidity stress testing sandbox modeling behavioral deposit runoff across insured/uninsured segments, scenario-driven cash-flow projection with survival horizon, and LCR/NSFR reconciliation on a validated DuckDB data layer with LLM-drafted committee narratives.
End-to-end AI trading system generating structured trade ideas using LLMs with JSON-based decision schemas, portfolio risk rules, and paper-trading execution via Alpaca API.
Interactive dashboard simulating multi-quarter capital projections under Base, Moderate, and Severe stress scenarios with RWA analytics and automated PDF/PowerPoint export.
Full analytics framework modeling market-linked insurance guarantees using GBM, Black-Scholes/Monte Carlo engines, delta-hedging backtests, and stress-testing modules.
ML-powered application detecting anomalies in foreign currency exposures from ERP data using Isolation Forest, enabling FX hedge audit and validation workflows.
PD models using Logistic Regression and XGBoost with Expected Loss computation, risk-band segmentation, and real-time applicant scoring dashboard.
Live option pricing application using Black-Scholes with Yahoo Finance API integration, annualized volatility computation, and interactive pricing interface.
Open to opportunities in quantitative risk, financial engineering, and analytics. Let's connect.