Examples of what I build and run.
Key-rate shocks through AFS fair value to AOCI to CET1, with a calibrated positive control. Capital headroom across CET1, SLR and TLAC for the largest US banks.
Python, Treasury curves, FFIEC, XBRL
Nulls matched to each signal's persistence, multiple-testing bars, known-answer securities and pre-registered verdicts before anything is believed.
Python, bootstrap, FDR
Self-contained research workflows pairing methodology, data-hygiene checks and runnable scripts so agents can assist with fundamental and quantitative analysis without silent errors.
Python, SQL
Every fact carries the date it became knowable, as-reported and restated kept apart. A security master that survives renames, splits and delistings, so nothing joins on ticker.
Python, DuckDB, SEC EDGAR, FFIEC
Staged jobs an agent can run and rerun safely, with resumable stages and machine-readable results. An independent watchdog checks freshness instead of trusting the jobs.
Python, SQLite
Call reports from 2001, holding-company XBRL and prices joined point-in-time. The same method runs on insurance, REITs and semiconductors.
Python, DuckDB, FFIEC CDR, SEC XBRL
A constrained fund shelf mapped to proxies and sleeves, with mandate-aware allocation, risk and scenario views, and report-ready output. Try it under Demos.
Python, Flask
Anonymized. No employer names, positions or returns.