Skip to content
Andres Carreon

NowSDET and AI automation at a payments fintech

I break software on purpose.

I'm Andres Carreon, a senior SDET who writes a lot of code. For 9+ years I've tested iOS, Android, desktop and web apps so they ship without surprises. Lately I build AI agents and the evals that keep them honest. After hours I ship my own products, from iOS apps on the App Store to tools that let AI agents use a website safely.

A recorded jevvium run. Goal: A registered user logs in with a valid email and password and is told they are logged in. Four decisions: tap Login, type the email and password, tap LOGIN, goal reached. Passed in 5.2 seconds, and the replay with plain Appium passed too.

Real run · jev-1.13.0 · iOS 26.5 simulatorSource
9+ years
9+yearstesting software, from enterprise to crypto to fintech
3 iOS apps
3iOS appsbuilt and shipped in React Native this year, two already on the App Store
40K minutes
40Kminutesof CI saved every month with self-hosted runners I set up
1st place
1stplaceglobal, at the AI Tinkerers Generative UI hackathon, with Capturia

Now building · open source

jevvium turns a test plan into an Appium test.

You write the steps in plain words, the way you would in a ticket. A decision model follows them through the app one tap at a time, and jevvium writes the path down as a plain WebdriverIO test that CI runs with no AI in the loop. On simulators or a real iPhone.

Read the code on GitHub

1Write the plan

The steps and what should be on screen when it works, written the way you would in a pull request.

# Add the bike light to the cart

1. Scroll to the Sauce Labs Bike Light and open it
2. Add it to the cart, then open the Cart tab

Expected: "My Cart"

2Explore once

A decision model picks the next action from elements that really exist on screen, and stops to escalate when it isn't sure.

  • tap "Home"
  • tap "Webview"
  • tap "Login"chosen · 0.87
  • tap "Forms"
  • tap "Swipe"
  • tap "Drag"

3Run forever

The path becomes a normal test. jevvium replays it once with plain Appium to prove it passes, then CI owns it.

$ npm run test:generated:ios
✓ login
✓ login-invalid-email
✓ forms-switch
no model calls at run time

47.7s → 16.3s

to explore four criteria, after rebuilding how it reads, decides and taps

~25 ms

per simulator tap through a small native helper, instead of about 400 ms

< $0.001

of model cost for those four criteria

4 at once

simulators exploring in parallel, one Appium server each

Field note: it found a real bug

In Sauce Labs' demo shop, “price, low to high” puts $7.99 after $49.99.

The app compares prices as text. The model couldn't see prices, so it wasn't sure the goal was reached. The expect check is what caught it.

Shipped lately

What I shipped lately.

Release notes from my own projects, newest first. Most of it built with AI agents I direct and review.

  1. 17 Nov 2026npm 1.4

    Bubblio Desk

    One line of middleware lets AI agents use any website: it learns actions from real agent traffic, and the owner allows each one before it goes live.

    Try the demo shop
  2. 16 Nov 2026iOS in review

    Booksic, rebuilt

    Rebuilt from scratch in two days: one Expo codebase for web and iOS, music composed by AI, in-app purchases and Stripe.

    Open Booksic
  3. 13 Nov 2026iOS 1.0

    Stylica and Pixalia on the App Store

    Native iOS apps in React Native for both products, with purchases verified against Apple and failing closed.

    See it on the App Store
  4. 11 Nov 2026open source

    jevvium 0.4

    Tests real iPhones and React Native apps, not only simulators, and waits for the app the way a person would.

    Read the code
  5. 11 Oct 2026hackathon

    AgentDesk

    A front desk for AI agents, built in a day with two teammates at the MongoDB hackathon in San Francisco.

    Read the code

Day job

9+ years of making software behave.

Quality engineering at a global consultancy, a crypto wallet used by millions, and now a fintech startup. Here it is as release notes.

  1. v3currentAug 2026 to now

    Fintech startup

    Software engineer, quality and AI automation

    name under NDA

    • Run an AI agent in Slack for every engineer, now the team’s first stop for alerts and outages. Each agent is isolated, never sees a credential, and needs its owner’s approval to act.
    • Built the eval suite and knowledge base for the customer-support AI agent, now at a 94% pass rate. A twice-daily LLM-as-judge review found a third of conversations were cut mid-session, a platform bug inflating every dashboard.
    • Built the end-to-end testing and the pre-merge gates for the core services, with mutation checks that prove the tests would catch a break on the money paths. The first run found 14 real defects.
    • Built the team’s shared Claude Code skills library, synced to every engineer’s agent, so everyone works with the same practices.
    • Set up self-hosted GitHub Actions runners that roughly halved backend test time and save about 40,000 CI minutes a month.
    • Ship production code on the backend and mobile app: KYC validations, a payment idempotency fix I traced with a repro harness, and the backend migration to AWS with a one-command rollback.
  2. v2May 2022 to Aug 2026

    Crypto wallet

    SDET, quality engineering

    self-custody, used by millions

    • Built the end-to-end automation from zero across mobile, desktop and the browser extension, focused on where users lose money or trust: send, receive, swap and buy. Manual regression stopped being the release bottleneck.
    • Built an AI bot that reads each pull request, decides what to test and opens a PR with the new tests; an alert named the pull request that broke staging.
    • Made failures explain themselves: Slack run reports with AI triage that tells a product bug from an environment problem.
    • Led four QA engineers and owned release sign-off.
  3. v12017 to May 2022

    Accenture

    QA engineer, test automation

    enterprise systems

    • Led a QA team of 15 building automated regression suites and page-object frameworks in CI.

After hours

Products I built on my own time.

Two with paying customers and native iOS apps on the App Store, one that lets AI agents use websites, and a hackathon winner. Building them is where most of my AI tooling comes from, including the agent that triages my own bugs.

Plus open-source tools, two hackathon projects and a few experiments.

All projects

Writing

Notes worth keeping.

Things I found interesting enough to write down: AI, attention, and building things alone. Most of them in English and Spanish.