SELECTED ENGINEERING WORK / 2026

Useful systems.
Measured outcomes.

Three working applications built around a practical question: can a system turn complex public data into something a person can verify, understand, and act on?

03Working applications
USAPublic data
PHP 8.2Deployable
01 / PROJECT INDEX

Built to be opened.
Not just looked at.

Each project has a working interface, dataset provenance, a measurable workflow, and documented limitations. Open one and try it.

Screenshot of the FilingIQ working interface
01 / DATA ENGINEERING / FINANCIAL RESEARCH

FilingIQ

Financial filing intelligence

Explore sourced financial figures from U.S. public company filings, compare annual trends, and export reproducible results.

PHP 8.2SEC XBRLJavaScriptREST APIs
Screenshot of the GridCast working interface
02 / MACHINE LEARNING / TIME SERIES

GridCast

Electricity demand forecasting

Train an interpretable ridge model on U.S. electricity observations, test it chronologically and compare against a seasonal baseline.

Pythonscikit-learnRidge regressionJavaScript

02 / HOW I BUILD

Engineering that survives
a closer look.

A good interface helps you discover the answer. A good system lets you see where it came from.

01

Source matters

Public data is cited; synthetic fixtures are identified; and out-of-date snapshots are labelled.

02

Baselines first

ML results are measured against simple alternatives instead of relying on model buzzwords.

03

Graceful failure

APIs can rate-limit or go down. The app communicates what happened and offers a useful fallback.

04

Runnable source

Each implementation includes automated tests and a documented local setup path.