I build software, then measure whether it actually works.

I take products, automation, and audits from an empty folder to a live release, then test what I built against real results. Every claim below links to the work or evidence behind it, so you can check it instead of taking my word for it. I'm bilingual in English and Spanish and currently studying cybersecurity.

WorkKindStatus
AME NetworksProduct, co-founderIn testing
VerdahProduct, founderRetired
Detection accuracy auditMeasured evaluationPublished
Oye MilaProduct, founderIn development
AI-personalized expert programsAutomationDeployed
Academic operations systemPersonal automationRunning
Local business digital presenceClient workLive
Browser extension policy auditAudit toolingResults private
Three interactive browser toolsToolsPlayable

Featured work

Demo agent website generated by AME

An automated marketing platform for independent insurance agents. An agent completes onboarding, and the system turns their information into a unique website, coordinated social assets, and ongoing marketing, while routing incoming leads into their CRM.

I lead the technical side: system architecture, automation, AI integration, infrastructure, payments, and production monitoring. The platform currently includes 13+ automated workflows and is designed to operate with minimal manual production.

  • Case studyHow it works, what I built, and results so far
  • Demo siteA fictional agent's site, generated by the platform
Verdah scoring a LinkedIn feed, names blurred

A Chrome extension that analyzed LinkedIn posts as you scrolled and labeled each one AI, Human, or Mixed, with a confidence level and explanation.

I built and operated the full product: browser extension, backend, billing and licensing, production monitoring, and release management. Model credentials stayed server-side and never reached the browser. It passed Chrome Web Store review and went through four production release cycles.

I retired it in September 2026 and shut it down cleanly: subscriptions cancelled, listing unpublished, infrastructure disabled, and credentials revoked.

  • ScreenshotAI, Human, and Mixed verdicts with confidence bands. Names blurred.

I hand-labeled 100 LinkedIn posts, 50 written by AI and 50 by people, and tested them against the detector I built. In this sample, 21 of 21 posts it classified as AI were actually AI, and 0 of 50 human posts were falsely flagged. It also missed 28 of the 50 AI posts.

I published the full confusion matrix and a threshold sweep showing why I chose not to trade the first false accusation for one additional detection.

Oye Mila product site

A Spanish-speaking voice companion for older adults who live alone. It's designed to work without an app, home Wi-Fi setup, or camera, and combines everyday conversation and reminders with updates for family.

I'm building the device software, backend services, scheduled check-ins and alerts, family portal, and bilingual product site, including the cellular-first setup that lets Mila work without home Wi-Fi.

Automation and client work

An automated system that turns an expert's existing methodology into a personalized digital product. Customers go through the expert's framework and receive an individualized deliverable and action plan, while fulfillment, follow-up, and the path into higher-tier services run automatically.

Built from concept to working product in 2 weeks.

An AI-assisted workflow that audits four online courses across multiple platforms, independently reconstructs grades, identifies recoverable work, tracks open instructor issues, and delivers one daily status report.

It recovered 150+ points through accepted late work, resubmissions, and grading corrections, and replaced manual checking across multiple platforms and inboxes.

  • Case studyAudit, triage, troubleshooting, and the daily brief

Took an owner-operated service business from no usable web presence and duplicate outdated profiles to a live website, branded email, and Google Business Profile.

Before changing the existing domain and DNS, I audited the services and accounts that depended on them, so the migration didn't disrupt the owner's existing tools.

A tool that audited 356 Chrome extensions by inspecting their actual manifest data rather than store descriptions, then evaluated the findings against Chrome's user-data requirements. It identified 12 extensions with missing or potentially inadequate privacy-policy coverage.

The findings involve real companies, so I don't publish them. I can walk through the methodology and redacted examples on request.

Tools

Readiness
ReadinessSecurity+ exam readinessOpen Readiness
R-Dojo
R-DojoChart-reading decisionsOpen R-Dojo
Thread
ThreadSales conversation practiceOpen Thread

Three self-contained browser tools built to run entirely client-side: no accounts, backend, or external database. Each is immediately playable, and progress stays in the browser.

What I won't claim