Work
Research systems, valuation models and decision tools built for analysis.
Tracing an M&A Shock Across Suppliers, Customers, Creditors and Competitors
Python + FastAPI
Models an M&A or financing transaction as a shock through supplier, customer, competitor, credit, market, ownership and technology relationships. The model evaluates base, upside and downside branches through DSCR, leverage, IRR, MOIC, covenant breach and composite risk diagnostics.
The reference network contains 12 entities and 24 directed relationships. A FastAPI and React application adds SEC XBRL and market data ingestion, role based access, PDF scenario memos, reconciliation checks, five historical calibration cases and a SHA-256 hash linked run ledger. Every assumption, propagation step and output can be replayed.
Deal Value and Debt Capacity Under Revenue, Rate, FX and Regulatory Shocks
Python + FastAPI
Recomputes DCF, debt schedules, DSCR, peak leverage, IRR, NPV and equity value under correlated shocks to revenue growth, EBITDA margin, WACC, interest rates, synergies, FX and regulatory timing.
Each run generates 5 to 30 seeded alternative scenarios, normalizes scenario probabilities and calculates portfolio P05 and CVaR05 loss. The FastAPI and React application records every branch in a SHA-256 hash linked audit trail and produces investment committee PDFs, valuation paths, scenario matrices and dependency graphs. SQLite is the default store with optional Supabase persistence and n8n workflows.
AI Trade Research with Fixed Execution Limits and a Tamper-Evident Audit Trail
FastAPI + React
Generates market diagnostics, investment memos and proposed trades through an AI research layer. A separate rule layer decides whether any order can enter paper execution.
The controls enforce symbol allowlists, restricted symbol rules, single name notional limits, order rate limits and an immediate compliance kill switch. Critical events are recorded in a verifiable SHA-256 hash chain. FastAPI, React, PostgreSQL, Redis, Docker Compose and n8n support the control plane, interface and scheduled workflows. Execution remains paper only. No model generated instruction can bypass the fixed rules.
Cross Asset Research from Market Regimes to SEC Filings and DCF Valuation
Node.js + Express
Combines cross asset price monitoring, the US Treasury yield curve, CBOE VIX regime diagnostics, Fama-French factor data, IPO tracking, SEC EDGAR filings, XBRL fundamentals, OHLC charting, probabilistic price scenarios, value screening and a DCF scenario lab.
Node.js and Express APIs support authenticated watchlists, deal memos, rule based alerts, PDF investment memos and document analysis. Each research surface identifies its data source across Yahoo Finance, Stooq, the US Treasury, CBOE, SEC EDGAR and the Ken French Data Library. The application runs on Cloudflare Pages with serverless API routing and persistent workspace storage.
From SEC XBRL Facts to Trading Comps, DCF Models and Analyst Packs
Python + SEC EDGAR
Pulls SEC EDGAR Company Facts and maps heterogeneous US-GAAP XBRL tags into standardized financial statements. The Python workflow calculates trailing twelve month KPIs and trading multiples including EV to Revenue, EV to EBITDA and P to E.
Each ticker run produces an Excel model with historical statements, projections, DCF valuation and sensitivity tables, plus a PDF analyst pack and structured JSON. Rate limited SEC requests, explicit mapping files, a versioned SQL data model and repeatable peer inputs preserve the path from filing facts to valuation outputs.
Momentum Tested Across Six ETFs and Rolling Market Windows
Python + Pandas
Runs a rolling momentum research workflow across SPY, QQQ, IWM, EFA, TLT and GLD. Each window selects a lookback from the training sample, applies the selected rule to the following test window and records parameters, trades, equity, drawdown, monthly returns and performance metrics.
Python, Pandas and NumPy power the calculations behind a FastAPI run service and Next.js review interface. Docker Compose supports repeatable execution. Every run receives a separate artifact directory containing parameter files, window logs, trade history, KPIs and charts. The outputs are research results rather than live trading performance.
Valuation Distributions and Reverse DCF for Implied Growth, Margin and ROIC
Python + Monte Carlo
Runs a ticker agnostic, driver based unlevered DCF with CAPM derived cost of equity and WACC. The Python implementation models revenue growth, EBIT margin and ROIC with triangular distributions and reports enterprise value, equity value and per share percentiles.
A reverse DCF solver back-solves the growth, margin or ROIC required to match a target enterprise value or share price. A 2,000 path PTRN demonstration produced enterprise value percentiles of USD 0.85 billion, USD 1.00 billion and USD 1.16 billion at P25, P50 and P75. Outputs include simulation CSVs, sensitivity tables, a valuation histogram and a one page methodology memo.
LBO Capital Stack Design Under Coverage, FCCR and Deleveraging Constraints
Python + Monte Carlo
Reconciles Sources and Uses, sizes a multi-tranche capital structure across TLB, second lien, mezzanine and RCF facilities, and models OID, upfront fees, PIK, mandatory amortization, cash sweeps and exit assumptions. Annual coverage, FCCR and deleveraging gates test whether the debt structure remains financeable.
A grid optimizer searches capital structure and exit configurations. Seeded Monte Carlo analysis then measures IRR, MOIC and credit robustness under rate and exit shocks. In the illustrative base case, net leverage falls from 5.88x to 3.99x by Year 3, minimum coverage is 1.95x, minimum FCCR is 1.52x, IRR is 20.7 percent and MOIC is 2.56x.