Selected projects
Credit risk frameworks, valuation and financial models, and quantitative finance work. Built independently in Python and Excel, benchmarked against regulatory and market standards.
Full Expected Loss framework on the Lending Club dataset (460,000+ records). WoE/IV pipeline selecting 17 features from 74. Interpretable scorecard scaled 300–850. Validated out-of-sample using Gini, KS, and CAP curve diagnostics against Basel III IRB standards.
Production-style EL framework across 307,511 applications and 5 relational tables. Two-stage LGD model. CCF-based EAD. IFRS 9 staging applied portfolio-wide. Basel III IRB validated — Gini of 0.481 clears the 0.35 IRB minimum.
Buy-side analyst simulation evaluating an $18.5B enterprise value acquisition target. Built an M&A target company profile covering enterprise value bridge, business overview, and leadership; a 5-year income statement projection model; and a trading comparables analysis. Findings synthesised into an investor-update report for the deal team.
A technically correct PD pipeline built to fail — intentionally. Documented the experimental design flaw at the root: a self-referential target variable that made meaningful discrimination impossible. The lesson: pipeline correctness and experimental validity are independent. Directly relevant to SR 11-7 and EBA model risk governance.
Multi-step binomial tree framework using backward induction and risk-neutral valuation. American and European options priced using live Nifty 50 market data. NetworkX used to visualise the full pricing lattice.
Python implementation of Black-Scholes using dynamically estimated annualised volatility from historical returns — replacing fixed-assumption approaches with a data-driven method tied to live Nifty 50 data.