ag/ Get in touch

Andrea Gubellini

Head of Quality Engineering

Treviso, Italy

agents at work on this page · scroll anytime

Animated demo: an AI orchestrator spawns agents that build a miniature copy of this website, test it, find a button padding bug and fix it.

01 — What I do

Four ways I put AI to work

15 years in tech, 10+ in quality engineering. I lead distributed quality organisations, turn QA gatekeeping into quality assistance, build automation that becomes the engineering standard, and put AI at the centre of it all.

01

AI testing agents

I design and build crews of autonomous AI agents that take a test case all the way: trace the flow, verify it in a real browser, write the test, prove it stable, review it and open the PR. No human in the loop until review.

  1. test case
  2. trace the flow
  3. verify in browser
  4. write the test
  5. run 10×
  6. review & PR
specialised agents per crew
~10
stability runs before every PR
10×
flaky retries tolerated
0

02

AI orchestration

I build AI release pipeline assistants: every failed release gets an AI first responder that reruns tests, correlates logs, screenshots, diffs and traces, names the root cause and the owning team, and calls go or no-go. Trusted to ship on its own.

  1. failed release
  2. rerun & correlate
  3. read the evidence
  4. root cause
  5. go / no-go
failed releases triaged by AI
100%
a year back to engineers
500+ h
precision before autonomy
≥95%

03

AI-augmented quality

Release quality on autopilot: automated merge, release and flakiness gates, with AI triage stepping in the moment one fails. Manual sign-off becomes the exception, not the process.

  1. pull request
  2. merge gate
  3. flakiness gate
  4. AI triage on fail
  5. release
releases needing manual sign-off
½
automated tests
+55%
stability bar to release
≥95%

04

AI-augmented management

I lead remote, multi-country quality organisations with AI in the loop: AI tooling rolled out to the whole team, hands-on workshops, and automated quality reporting for leadership.

  1. AI tooling
  2. hands-on workshops
  3. team adoption
  4. automated reporting
quality engineers led
~20
countries
7+
time zones
3

02 — Impact

Numbers, not adjectives

Fifteen years in tech, ten in quality. Here's what that adds up to when you put AI to work.

  • 15years in tech, 10+ in quality engineering
  • ~100engineers building on my teams and frameworks
  • 10+AI agents designed and built
  • 500+engineering hours a year handed back by AI
  • 3×faster test suites
  • –34%critical production bugs

a.

Autonomous AI testing agents

Crews of AI agents I design and build turn a plain test case into a stable, reviewed end-to-end test and open the PR themselves. They even surface real product defects, like concurrency bugs humans missed.

agents per crew
~10
stability runs per test
10×
flaky retries tolerated
0

b.

AI release triage

Every failed release gets an AI first responder. It reruns, reads the logs, traces and diffs, names the root cause and the owning team, and calls go or no-go. Trusted to approve releases on its own, one precision gate at a time.

failed releases triaged
100%
a year back to engineers
500+ h
precision before autonomy
≥95%

c.

Test automation at scale

I build test-automation frameworks that become the engineering standard. Flaky turns stable, slow turns fast, and developers start writing the tests themselves.

end-to-end stability, from ~66%
90%
faster suites
3×
engineer contributors
~80
commits
~2,000

d.

Quality as a system

Shift-left, automated release gates and coverage bars agreed with engineers, not imposed on them. Quality stopped being a gate at the end and became how the team ships.

critical production bugs
–34%
coverage bar in 12 months
2×
manual sign-offs
½
automated tests
+55%

03 — Experience

From passion to Head of Quality

10+ years in QA: from test analyst to leading Quality across all 4 product tribes at TheFork.

  1. 2023 – present

    TheFork · an American Express company

    Head of Quality Engineering May 2025 – present

    • Took on leadership of Quality across all 4 product tribes. The organisation grew to ~18–20 SDETs and Quality Analysts covering B2C web, iOS/Android apps, the B2B restaurant platform (web, mobile, tablet), APIs and internal tools. It is a fully distributed, remote-first team spread across 7+ countries (France, Spain, Portugal, Italy, Hungary, Ukraine, Poland) and 3 time zones.
    • Line-manage SDETs and Quality Analysts: hiring, performance and career paths.
    • Own quality strategy, OKRs and KPIs for engineering. Run a monthly quality review and report quality health to engineering leadership.
    • Own the team and contractor budgets and manage vendor partners end to end: contractor partners (e.g. EPAM) for staffing and delivery, and tooling vendors (e.g. SonarQube, BrowserStack) for contracts.
    • Own quality incident response: on-call bug triage and release-duty rotations, root cause analysis, war rooms and post-mortems (Rootly).
    • Moved release quality from manual sign-off to automated gates. Removed manual QA validation from production releases and introduced an E2E merge gate on PRs, a mobile release gate (≥95% stability) and a 10-run, zero-retry flakiness gate for test code. Releases needing manual QA approval halved (~12% → ~6%), and automated web and API tests grew by ~55%.
    • Sponsored and owned an AI release assistant that triages every failed release run. It reruns failed tests, matches history, and analyses logs, screenshots, PR diffs and APM traces. It then posts the root cause, owning team and a go/no-go recommendation. It won leadership approval for autonomous approvals, each step gated on ≥95% precision, and saves the SDET team 10+ hours of triage per week.
    • Designed and built, hands-on, autonomous AI testing agents (~9 agents) that automate a test case end to end. They trace the flow across code, docs and Datadog traces, verify it in the browser, write the Playwright test, run it 10× for stability, review it and open the PR. The first pilot went from backlog to PR in 1 day and found a real product concurrency defect.
    • Drove AI adoption across the Quality team: AI tooling, hands-on workshops and automated quality reporting.

    QA Engineering Manager Jan 2023 – May 2025

    • Authored the Quality Engineering vision (governance-approved, 2023). It moved the company from QA gatekeeping to quality assistance, with developers writing tests on SDET-built frameworks.
    • Designed a shift-left process (Definition of Ready/Done, test planning during shaping). Piloted it with 3 teams (86% of analyses met Definition of Ready) and rolled it out to all product tribes in Feb 2025.
    • Founded the Playwright + TypeScript framework and led the migration from legacy Selenium and API suites. E2E stability went from ~66% to ~90% and suite time from ~12 to ~4 minutes. The framework became the engineering standard (~80 contributors, ~2,000 commits).
    • Owned the code-quality gate programme. Doubled the minimum coverage threshold (34% → 66%) in 12 months, setting thresholds with engineers through a company-wide survey.
    • P0/P1 production bugs fell ~34% company-wide between 2023 and 2025.
    • Launched a weekly Quality & Release digest for all of engineering. Presented quality strategy at Engineering Forums and to engineering directors.
    • Took over a contracted automation team (~3–5 people) and grew it into an in-house SDET team. Hired ~6–8 SDETs and Quality Analysts, wrote the role descriptions and revised the career paths.

  2. 2021 – 2023

    Prima

    QA Engineering Lead 2021 – 2023

    • Defined the QA Engineer role, its expectations and the interview process. Led the QA community of practice.
    • Coached Scrum teams on quality assistance and shift-left testing.
    • Built automation frameworks and libraries for microservices (Python, JavaScript, GraphQL, REST, Robot Framework). Set up CI/CD and ChatOps integrations.

  3. 2020 – 2021

    CHILI

    Senior QA Engineer 2020 – 2021

    • Designed a new scalable test automation framework (Robot Framework + Python) with Jenkins CI and ChatOps.

  4. 2018 – 2020

    Facile.it

    Senior QA Automation Engineer 2018 – 2020

    • Led automation activities. Built API and front-end test suites in GitLab CI, custom Python libraries and Dockerised test images.

  5. 2011 – 2018

    Earlier roles

    QA Specialist, Placebuzz (2017–2018)Lead Test Analyst and Test Analyst, Tata Consultancy Services (2016–2017; On the Spot Award, 2016)IT & SAP Helpdesk Consultant, LAMP (2011–2015)

04 — Skills & tools

What I work with

The short list. Leadership first: agents still need someone to tell them what good looks like.

01 Leadership

  • Quality strategy
  • Influence without authority
  • Remote, distributed teams
  • Vendor and budget management
  • OKRs and KPIs
  • Hiring and career paths

02 AI and quality

  • AI testing agents
  • AI release triage
  • Risk-based testing
  • Shift-left
  • Merge and release gates
  • Incident response (on-call, RCA, post-mortems)

03 Tech

  • Playwright
  • TypeScript
  • Node.js
  • REST and GraphQL
  • Python
  • Robot Framework

Tools along the way

  • Playwright
  • TypeScript
  • Selenium
  • Datadog
  • SonarQube
  • BrowserStack
  • Rootly
  • Jenkins
  • GitLab CI
  • Docker
  • Robot Framework
  • Python
  • JavaScript
  • GraphQL
  • REST

Languages

  • English · full professional, lived in the UK
  • Italian · native
  • Spanish · basic
  • Russian · basic

Teaching and community

  • Robot Framework course instructor 2024–2025
  • Course instructor, GPI 2017–2018
  • Robot Framework open-source contributor 2019–2021

05 — Contact

Let's talk agents

Building AI agents for testing, or a quality team that works alongside them? Let's compare notes.

andreagubellini@gmail.com · github.com/andreagubellini · Treviso, Italy