Skip to content

Case studies

Each one follows the same memo: context, problem, approach, what I built, outcome, tools — and where else the technique applies.

01 · Funnel · Policy · SQL

Lifting portfolio qualification from 30% to 43%

30% → 43% portfolio qualification rate

Six lending partners, 100K+ monthly borrowers, and a funnel nobody could see inside. Rejection-code analysis turned "we lose most applicants somewhere" into a ranked list of fixable constraints.

read_case_study
02 · Policy · SQL · Tooling

Policy as code: 112 rules, 200+ variants, six lenders

200+ policy variants configured

Six partners, six written credit policies, one rules engine. Turning prose policy documents into versioned, testable, executable configuration — and keeping them honest with an automated regression harness.

read_case_study
03 · Tooling · Policy

PolicyLens: making rule-engine policies readable — and editable — by analysts

12,700+ lines of Python

Credit policies were stored as a table of rule identifiers plus a boolean expression referencing them — unreadable by eye. PolicyLens renders them as human logic, diffs two policies order-insensitively, and round-trips edits back into deployable configuration.

read_case_study
04 · Risk · Dashboard · SQL

Cohort and vintage delinquency monitoring for a US bank

+25–30% delinquency prediction accuracy

Concurrent and ever-delinquency matrices, vintage curves and early-warning indicators for Axos Bank — turning a monthly regulatory reporting grind into a portfolio view that surfaces deterioration while it is still early.

read_case_study
05 · SQL · Tooling · Dashboard

Removing 80% of manual vendor risk reviews with an ETL pipeline

80% of manual reviews eliminated

Third-party vendor risk assessment was a recurring manual compliance workload. A Python pipeline against the vendor-risk API turned it into monitored data — with the exceptions, not the whole portfolio, reaching a human.

read_case_study
06 · Dashboard

Bootstrapping a bakery on SKU-margin analytics

SKU-level margin tracking

The Oven Vibe is a D2C bakery I built and run. Same analytical method as the lending work, applied to a business where I am also the one who has to live with the decisions.

read_case_study