Felipe Knorr Kuhn, quality engineer in the San Francisco Bay Area

Quality engineer with two decades of finding how software and hardware fail and making that failure cheaper to discover. Web, mobile, hardware, and systems used by millions. These days I’m all in on AI: building games with it, mapping where it shines and where it breaks, and pointing it at the open source projects people depend on.

Open to work Quality engineering, test infrastructure, and AI evaluation roles.

felipe.spec.ts

describe('Felipe')

  1. has shipped at global scale Yahoo!, Netflix, mempool.space
  2. tests across mobile, hardware, and web 6 domains covered
  3. builds the test infrastructure, not just the tests 334 merged PRs at The Mempool Open Source Project
  4. uses AI as an instrument, not an authority OpenSats Red Team grant
  5. knows Bitcoin from explorer to signing device mempool, bitcoinjs, Zeus, SeedSigner
  6. still plays games press start below

6 passing

Coverage report

Most careers go deep in one kind of system. Mine kept crossing over. Pick a domain to see the work that covers it.

Uncovered domains: 0

Work

From Porto Alegre to Netflix, two decades with the same mission under every title: find out how software fails before users do, and build the tests, tooling, and habits that keep it from failing again, sharing them with the teams and open source communities that depend on it.

30.03° S, 51.22° W
  1. 2003 – 2009

    Early career

    Porto Alegre, Brazil

    Various roles across Porto Alegre’s software industry

    Six years in a range of roles that taught me how software gets built, before I specialized in how it breaks.

  2. 2009 – 2011

    Brazilian federal government

    Porto Alegre, Brazil

    An agile quality group for citizen-facing systems

    I was part of a dedicated agile quality group, testing and automating systems built for Brazilian citizens, from the Presidential Office to the tax collection department and the Ministry of Labor, all on open source technology.

    • Web test automation with Selenium, TestNG, and ReportNG
    • Spoke at Agile Denver (2010) and gave a CAST 2011 lightning talk, "Bug or feature: the importance of being context driven"
    • Published articles on test automation in testing magazines
  3. 2011 – 2016

    Yahoo!

    São Paulo, then Sunnyvale, California

    From Windows 8 apps to a rebuilt Yahoo Search

    In São Paulo I worked on Yahoo Profile and the Windows 8 Messenger and Mail apps. In Sunnyvale I moved to the native iOS and Android apps, then to the complete reboot of the Yahoo Search stack.

    • Built the first automation framework for the new Windows 8 Metro apps, open sourced as RainbowDriver, with a continuous integration pipeline running on real devices
    • Reinvented how Yahoo Search tested and automated result evaluation for the new knowledge-card results and A/B experiments
    • Q4 2011 “You Rock!” award with the Yahoo Profile team
    • 2013 “Super Star Team Award” with the Yahoo Mail Mobile team, for the Quattro launch: one global release across Android, iOS, Windows 8, and desktop
  4. 2016 – 2017

    Tile

    San Mateo, California

    Bridging the whole company, from firmware to customer support

    As principal quality engineer, reporting to the VP of Engineering and then to the CEO, I worked across the entire stack: from the trackers’ firmware to triaging customer support bugs and following them through with the development teams.

    • Shipped the Tile Mate, Tile Slim, and Tile Pro products
    • Built physical automation for the Bluetooth trackers, and tested features like Smart Alerts
    • Validated firmware fixes in extreme weather, and characterized device range in real-life scenarios during the initial hardware design phase
    • Inventor on US patent 11,968,321, “Task completion in a tracking device environment”: a scavenger hunt built from tracking devices
  5. 2017 – 2026

    Netflix

    Los Gatos, California

    From the TV video canvas to Netflix Live

    Senior software test engineer on the Netflix TV experience: first on the TVUI Video team, then TVUI Core, then as test lead for Live. Throughout, I worked closely with the Device Reliability, playback, and backend engineers to investigate and fix field issues on TV devices, on call and off.

    • TVUI Video: immersive and interactive experiences around the video canvas, including merchandising, previews, seamless post-play, UI audio, and updatable splash-screen animations
    • Special takeovers: Spookyflix for Halloween, the Roma UI treatment, the immersive experience for Martin Scorsese’s The Irishman, and the Lost in Space 3D billboard
    • Alexa voice integration, with an automation framework that used voice synthesis to measure end-to-end title launches and navigation
    • TVUI Core: modernized automation and tooling, and coached teams on quality and testing
    • Live: test lead across all platforms for the launch of Netflix Live and the features that followed, including real-time voting, live chapters, the NFL Christmas games, and inline UI that updates with a title’s schedule and state
    • Built the Live automation strategy around fake live streams, wrote hands-on automation for TV, and built infrastructure that unblocked every partner developing and testing real-time updates, voting, and chapters
    • Spoke at Applitools’ “Back to the Future of Testing” on how software testing keeps evolving

Open source

Public code I’ve shipped, reviewed, and red-teamed. Most of it is Bitcoin software that people trust with real money.

  1. Testing tools and frameworks

    Open source test automation

    Documentation, code snippets, reference frameworks, and community support

    I built test automation on open source tools like Selenium, Watir, TestNG, and ReportNG, and gave back with documentation, code snippets, reference frameworks other teams could start from, and answers for other testers in the community.

  2. 2021 →

    The Mempool Open Source Project

    mempool.space

    Quality and release infrastructure for mempool.space

    mempool.space is the Bitcoin explorer most of the ecosystem checks before it trusts a transaction. I built and ran the machinery that let a small team ship it safely: tests, CI, Docker images, and the packages people run on their own nodes.

    334 merged pull requests across 15 repositories

    • Laid the foundation for backend unit tests and grew the Cypress end-to-end suite across Bitcoin, Liquid, and mining views
    • Coverage reports on every pull request, and a way to run the e2e suite against any branch or PR
    • Multi-arch (amd64 + arm64) Docker builds, a Node version test matrix, and config validation before deploy
    • Hardened CI supply chain: actions pinned to full commit SHAs and an audit for workflow updates
    • Maintained the mempool app for Umbrel and StartOS, the home-node hardware many users run it on
    • CI and build work for electrs (Rust) and the transaction accelerator
  3. 2026 →

    OpenSats Red Team grant

    Independent research

    Hardening Bitcoin open source with AI

    OpenSats funds researchers who red-team critical Bitcoin software. My grant applies AI to finding defects in open source Bitcoin projects, each one scoped and disclosed through the project’s own channel.

    • Published the target-selection strategy and a scope for each project
    • An AI-driven loop that investigates issues and proposes hardening, refined on each project and carried to the next
    • Every finding ships with proof-of-concept code that triggers the vulnerability, plus a proposed fix or a pull request that hardens the project’s infrastructure
  4. 2026 →

    SeedSigner

    Air-gapped signing device

    Parser hardening and code review

    SeedSigner is a DIY, air-gapped Bitcoin signing device built on a Raspberry Pi. It is the first target of my Red Team work: I review pull requests and send fixes where the device trusts its input too much.

    • Opened a pull request that makes the SettingsQR, sign-message QR, and PSBT parsers fail closed on malformed input instead of crashing, with unit tests
    • Reviewing pull requests from other contributors
  5. 2026

    Zeus

    Bitcoin and Lightning wallet

    Private security report

    Zeus is an open-source, self-custodial Bitcoin and Lightning wallet for iOS and Android. I reported an issue to the maintainers privately. The details stay with them, which is how disclosure should work.

AI as an instrument, not an authority

Under an OpenSats Red Team grant I use models to look for defects in Bitcoin open source. The model widens coverage. Judgment stays human, and the model never gets to publish.

I build with it too: Layoff Simulator, Severance Dash, Open to Work, and Quick Sync, the Game Boy Color games further down this page, and FOMO Charts, a Bitcoin analytics site.

  1. 1 scope

    Pick a target I am allowed to study, where failure has a real cost.

  2. 2 review

    Use the model for coverage: read code against documented invariants.

  3. 3 triage

    Most output is noise. Only claims I can reproduce with a proof of concept survive.

  4. 4 disclose

    Reproduce on my own checkout, then report through the project’s channel.

Bitcoin

Five years contributing to The Mempool Open Source Project, and now red-teaming the software Bitcoin users trust with their keys. Below is the network right now, read from the explorer I helped ship.

Latest block
—
Next block fee
— sat/vB
Unconfirmed
—
Mempool size
—

Connecting to mempool.space…

Initial Red Team targets

Three layers, from the code wallets import to the device in your hand, so the work covers different kinds of failure instead of the same bug three times.

  1. Library

    bitcoinjs-lib

    Imported by many wallets, so one defect multiplies. PSBT, address encoding, and the glue between packages.

  2. Mobile wallet

    Zeus

    A self-custodial Bitcoin and Lightning wallet for iOS and Android. An issue went to the maintainers privately, and the details stay with them.

  3. Signing device

    SeedSigner

    An air-gapped Raspberry Pi. What the screen shows must match what gets signed.

Follow the work

The Red Team work is written up as it happens: how each target was chosen, what is in scope, and how findings get disclosed.

Games

I wanted to make games before I knew testing was a job. There was barely a games industry in Brazil then, so I tested everything else. Games taught me English, and they taught me to look for the glitch.

Now I make them. The first cartridges are my own Game Boy Color games, built with Codex and the ModRetro plugin, and drawn from recent experience: , , and . They are real ROMs, running in an emulator right on this page. Press Start.

The last cartridge is a rematch. In 2012 I wrote a Sikuli script to play Google’s Olympic hurdles doodle. It finished in 10.8 seconds with two stars, and it never learned to jump. This one does: play it yourself, or let the bot run.

Off screen, the hobby is hardware: Game Boy cartridges, the ModRetro Chromatic, and the Analogue Pocket, where I contributed to the Pupdate updater.

See more games

Layoff Simulator 1/5

Layoff Simulator Press start

knorrium BOY COLOR

You’re the new VP of Operations. The board wants payroll down 20%, one department gone, and the stock above $12. Decide who goes, role by role. Arrows or WASD to move, X or K for A, Z or J for B, Enter for Start, Shift for Select.

Side projects and contributions

Things I built because a tool was missing, and fixes I sent upstream because I was already in there.

  • 2026

    FOMO Charts

    Bitcoin investment analytics on live mempool.space data: a price chart, market insights, a tracker for missed opportunities, and a DCA simulator. Portfolios stay in the browser.

    • AI
    • Bitcoin
    • Web & CI
  • 2026

    treemap

    Breadcrumb navigation and UX fixes for a macOS disk-usage visualizer.

    • Web & CI
  • 2023

    bitcoinjs regtest-server

    Added RPCAUTH support, address targets for block generation, and config checks to the regtest server bitcoinjs-lib tests against.

    • Bitcoin
    • Web & CI
  • 2023

    Pupdate

    Contributed to the Analogue Pocket updater so a fresh SD card gets its core directories created automatically.

    • Hardware
  • 2020

    Agile Testing Condensed, Brazilian Portuguese edition

    Janet Gregory and Lisa Crispin’s short introduction to agile testing, translated into Portuguese with Paulo Gonçalves for testers in Brazil.

    • Publications
  • 2015

    RapidReporter for Mac

    A Swift port of Shmuel Gershon’s note-taking tool for exploratory testing sessions.

    • Mobile
  • 2015

    iOS and Android apps

    On the Map, MemeMe, PitchPerfect, and Popular Movies, built through Udacity’s iOS and Android nanodegrees, plus a tvOS TVML experiment.

    • Mobile
  • 2012

    RainbowDriver

    A WebDriver server for Windows 8 Metro apps, built at Yahoo! for the Messenger and Mail apps. Selenium tests in any language could drive native apps through UI Automation.

    • Web & CI
  • 2012

    Google Books Downloader

    A PhantomJS and CasperJS proof of concept that saves free Google Books titles. 28 stars and counting.

    • Web & CI

Writing

Notes on software quality, Bitcoin, and the tools, AI included, that I use to probe both. I’ve been writing in public about testing since 2010.

All posts

Latest

Hiring for quality? Let’s talk.

I’m looking for my next team, in quality engineering, test infrastructure, or AI evaluation. I do my best work where reliability is the product, whether that’s a phone app, a device, or a system with millions of users.