Zoltán Balogh

Hands-on engineering manager. Software platforms, developer tools and quality engineering. Helsinki, Finland.

If you are building something serious, get in touch.

Coding professionally since 1992, leading technology and projects since 2004, managing people since 2012. I've worked on SMS centres, MRI scanners, anti-virus gateways and telecom auditing, then on four Linux platform families (Maemo/MeeGo, Ubuntu, automotive Linux and today SUSE Linux).

I still write code as a manager. At SUSE I develop tests, packages and tools, and at Canonical I had as many commits as the engineers on my team. Staying in the trenches with my engineers means I understand what they're up against, clear their blockers, and recognise good work when I see it.

My management style is Nordic. Kind, demanding, no BS, and big on listening. I build and grow teams that ship and have fun doing it.

Career

Click a row to see the product, the team, what I did, what I learned and a story or two.

2018 - now SUSE · Senior Manager, Quality Engineering I lead a globally distributed QE team for SUSE Linux and still work hands-on in openQA, packaging and tooling.
product
SUSE Linux Enterprise Server, openSUSE
team
10-19 engineers, product owners and scrum masters in China, Finland, Italy, Spain, Germany and Czechia
role
people manager and contributor

What we do

I joined to lead one of the Maintenance QA teams, validating maintenance updates before they reach SUSE Linux Enterprise customers. Later Maintenance QA merged into the QE department that owns the quality of SUSE's Linux platform. Today my team develops tests, validates maintenance updates and develops openQA, the open source test automation system behind openSUSE and SLE.

The team

The team grew from a small group to 10-19 people, with close to zero attrition. Some of my team mates were promoted. My job is hiring and developing people, removing impediments, resolving conflicts and keeping the atmosphere positive. Sometimes that is technical work, sometimes it is just listening.

Things I started

  • Squad rotation. Engineers move to another team for 2-3 months to break silos and single points of failure, share practices and grow.
  • PO on the bench. Each product owner sits down with the QE staff and talks about roadmap, achievements and pain points.
  • Writing in public. I started my tech blog in 2021 to show that it is easy. Many engineers in the department followed, and for them publishing is now normal.

My own contributions

  • zypper-changelog-plugin shows the changelog of a package before you install it. It ships in openSUSE and SLE.
  • zypper-rdepends-plugin lists recursive reverse dependencies, so we know what an update can break. It started as a Hackweek project and led to the --requires-pkg option in zypper itself.
  • Contributor to dozens of SLES and openSUSE packages. For several years I was the maintainer of vim in SLE 12 and SLE 15.

The AI shift

LLMs changed how we write tests and how we run openQA. For me the big deal is that we can finally pay back the technical debt nobody had time for, and improve the testability and test coverage of SLES.

  • I improved the performance of our tests and saved about 10% of the openQA machine time.
  • We measure the binary function coverage of SLES tests, so we can see which parts of a program our tests really exercise.
  • I enable upstream test suites in our packages, so that SLES is tested with the tests its developers wrote.
  • I research and experiment with retrieval augmented generation (RAG) on local LLMs. I built RAG systems over Bugzilla and Redmine data and over the system journal, to understand what causes pain to our customers and engineers. Everything runs locally and no data leaves the machine.

The ideas are all organic, human innovations. The implementation is driven by AI. I keep the old disciplines of test-driven development, planning, documentation and small, frequent releases. I do not commit code I do not understand, and I am accountable for every line I commit.

Helping the department change gears is part of my job too. I do not preach to the sceptics, I show them what I do. AI can finally build the ideas you never had time for and clean up the old code you always wanted to refactor.

What I learned

The most important part of a manager's job is that people are productive and have fun at the same time. The two go together.

openQA · Perl · Python/pytest · Go · OBS/osc · RPM · zypper · Salt · Grafana · eBPF · RAG · Ollama · ChromaDB

2017 - 2018 Comptel (acquired by Nokia) · R&D Manager I led a 23-person engineering team building service orchestration software for telecom operators.
dates
Dec 2017 - Jul 2018
product
cloud-native service orchestration and workflow management for telecom operators
team
23+ engineers in one Scrum team
role
people manager

The team

This was a pure people management job, and the most disciplined and most agile team of my career.

Textbooks say a Scrum team is 7 people. Ours was 23 and it worked. The daily stand-up never took more than 15 minutes, because everybody came prepared. People were disciplined and respected each other.

What I learned

When people leave a paper trail behind the work they have done, everybody sees what is going on and the estimates get better every sprint. Then the cost of engineering can be planned and measured.

Then SUSE called. It took me back to Linux and open source, where I wanted to be.

Scrum · Jira · telecom OSS

2017 Link Motion · Engineering Manager Automotive Linux, an infotainment and dashboard OS for cars, with an offline container-based SDK.
dates
Apr - Dec 2017
product
a GNU/Linux platform for connected cars, covering infotainment, entertainment and the dashboard
team
4 direct reports, in a small Finnish-Chinese company
role
engineering manager

When Ubuntu Phone ended, a few of my closest Canonical colleagues and I moved to Link Motion. Same job as before, covering the platform toolkit, developer tools, CI and automated QA. One highlight was a collaboration demo where our platform ran in a Lamborghini. That was cool.

What I built

I brought the container-based cross-build approach from Canonical to Link Motion, where it did not exist before. There I turned it into a cross-platform build and packaging framework that doubled as an offline SDK. Not rocket science, but it worked and I enjoyed it. Towards the end I got good results on the Yocto side, but the project ended before it could go further.

The tool outlived the company. I kept developing it and adapting it to new needs, and today it lives on as sysroot-tools for openSUSE and SLES (blog).

LXC · Yocto · Qt · cross-compilation · CI

2012 - 2017 Canonical · Engineering Manager, Ubuntu SDK and UI Toolkit I built the team behind the UI Toolkit and the SDK for Ubuntu Phone, and wrote code alongside them.
dates
Jan 2012 - Apr 2017
product
Ubuntu Touch (Ubuntu Phone and convergence), with the UI Toolkit every app was built on, the Ubuntu SDK IDE, the application framework and the platform APIs. Shipped on bq and Meizu phones.
team
about a dozen engineers, fully remote, many countries
role
engineering manager and product owner, hands-on

When the Nokia story ended I wanted to use what I had learned about building a mobile Linux platform. Canonical had just started Ubuntu Phone, still a secret project at the time. The goal was the same as MeeGo, but the teams were much smaller, more disciplined and faster.

There was no team when I joined. The UI Toolkit and SDK team was founded with me. I got a few engineers assigned and hired the rest. For the first half year I reported directly to Mark Shuttleworth, the most demanding boss I ever had. He sees through anything that is not backed by facts. From him I learned how to report effectively.

What I built

  • Qt 5 for Ubuntu. Packaging Qt 5 for Ubuntu, together with a colleague, was my biggest early contribution. It was new to Ubuntu, and the ARM port especially was a lot of work.
  • The test system of the UI Toolkit. End-to-end automatic tests, about 400 autopilot tests plus validation of every core and system app before each release. We released the toolkit twice a month.
  • The Ubuntu QtCreator plugin, the IDE side of the SDK.
  • The container-based cross-compiler SDK. This was the real innovation. At the time no Linux platform offered developers a container-based cross-compiling SDK. Building an app for a phone used to mean bootstrapping a 1.7 GB chroot for each target. We replaced that first with prebuilt images and then with LXD containers, about 400 MB images that start in seconds and build, package and run apps for any target. Blog posts

The commit history is public. I have about 700 commits in ubuntu-ui-toolkit and about 340 in qtcreator-plugin-ubuntu.

I also presented our work in public, for example the Ubuntu SDK roadmap discussion and a talk on the Ubuntu Touch UI Toolkit and SDK at a Hungarian Free Software Foundation event (in Hungarian).

Stories

The 20-minute browser. In a meeting with Mark we debated whether to build our own browser for the phone. I said we should. Mark said he would like to hear about it later. While the meeting went on I wrote a QML app with a web view, back and forward buttons and a URL bar with history. Twenty minutes later I turned my laptop towards Mark. With a good web engine underneath, the browser itself is the easy part. He laughed and was convinced. He appreciated code, demos and facts.

The blue component. One Friday a UI component turned blue in a release. I tested it, saw nothing broken and assumed the designers had changed their minds. On Saturday morning the messages started. The component should have been orange. A designer asked if I had not noticed how ugly it was. Honestly, no. I like beauty and art, but I am not a visual designer and my eye for colour should not be trusted. We fixed it, and I added colour checks to the tests. If something matters, test it.

What I learned

  • A dozen disciplined and highly motivated A+ engineers can do what a huge company did with 100+ equally talented ones. In software, less is often more.
  • At Ubuntu even senior managers were skilled engineers and hands-on contributors. Since then being hands-on is part of my identity as a manager.
  • Remote work, years before Covid. Watch, listen and observe, because there is always more that you do not see and hear. Be kind and demanding at the same time.
  • The toolkit versioning debate taught me that we are committed to stable APIs, not to stable code. Freezing code is a luxury of very mature products.

Ubuntu Phone was cancelled in April 2017. We had felt that leadership's focus was moving elsewhere. I did my best to help my team find new jobs.

Qt/QML · C++ · Python · QtCreator · LXC/LXD · Debian packaging · autopilot · CI · Launchpad/bzr

2008 - 2012 Nokia Devices, Maemo/MeeGo · Senior Project Manager, Project Coach Project manager of project managers for the developer offering of Nokia's Linux devices.
dates
Dec 2008 - Jan 2012
product
the Maemo and MeeGo developer story, everything a developer needed to build for Nokia's Linux phones and tablets
team
coached 7-9 project managers and scrum masters, 80+ engineers on multiple sites
role
project coach and product owner of projects worth 2-5 million euros

This is where the Linux platform chapter of my career started. Nokia wanted to break into the handheld market with an open source platform, and it did. About two million Linux devices ended up in the hands of non-technical people. The vibe was phenomenal.

All the developer offering projects were mine, including the Python and web runtimes, the native platform SDK, the application developer SDK, the platform APIs and software update. I am proudest of the native platform SDK. The N900 and N9 videos show products my teams were part of.

What I learned

  • How to run agile teams and projects at scale, and how to build roadmaps for multimillion euro projects.
  • Technical excellence needs a business case behind it. Since then I tie engineering decisions to business outcomes.

On 11 February 2011 I was in the big auditorium with everybody else when the "burning platform" strategy was announced. It was a very hard day. Maemo and MeeGo did not fail on technology. The business did not click.

I missed coding a lot in this job. Moving from code to project management was my choice, and it is the reason I made sure to get back to code later.

Agile/Scrum · Critical Chain · roadmapping · Maemo · MeeGo · Qt · SDKs

2004 - 2008 Nokia Networks / Nokia Siemens Networks · Software Architect, Technology Manager We invented and sold an automated health check service for live telecom networks. Coding met business.
dates
Nov 2004 - Nov 2008
product
RRHC, the Regular Remote Health Check service for Nokia network elements
team
grew from 4 to 10-12, plus local delivery engineers around the world
customers
50+ operators in Asia, Africa, the Americas and Europe, among them some of the biggest in the world
award
eXtreme Customer Delight Championship, Nokia Networks, 2007

An agent collected tons of data and monitored hundreds of metrics on live telecom systems. The system analysed the data against reference values, collected the findings and created a report. Then I presented the findings to the customer, with our suggestions on how to configure and operate the network to make it more stable and more profitable. Everybody in the team had developed one or more of these systems. I was the SMSC expert.

What I did

Wall to wall development of the agent in Perl. The hosts were HP-UX, SunOS and Linux, and back then Perl was the only portable interpreted language. The analytics were Perl and C, the UI PHP. And then I put on a suit and met the technical executives of big operators.

In a small team we did everything, from pricing, budgets and project management to technology management and coding. I calculated the cost of delivering the service. These services were expensive, and they made a lot of money for Nokia.

Before the Idol finale. We ran a preventive audit for an operator before their local Idol TV show, which meant a huge peak of SMS votes. The disk system that held the charging records, the data used to bill every one of those votes, had no RAID and no mirror. Incorrect high availability setups were very common. That is the kind of thing RRHC was for.

What I learned

This is where coding met business. A geek can wear a suit, sit with executives and talk money, as long as the facts come from the code. It was also my first distributed team, because the services were delivered with local engineers all over the world.

Perl · C · PHP · HP-UX · Solaris · Linux · SMSC · performance analysis · service business

2004 F-Secure · Senior Software Engineer Porting an internet anti-virus gateway from BSD to Linux.
dates
Feb - Nov 2004
product
F-Secure's anti-virus gateway for internet protocols

My job was to port the gateway from BSD to Linux, in C and C++. We tested on SUSE Linux because it was the easiest Linux to install and operate at the time. Fourteen years later I joined SUSE.

The team was really good, and I learned a lot about internet security, malware and viruses. It was short only because Nokia Networks made me an offer I could not miss.

C · C++ · FreeBSD · Linux · SUSE Linux · network protocols

2002 - 2004 Philips Medical Systems, MR Technologies · Software Design Engineer The operator console of an MRI scanner. Regulation is not a joke.
dates
Mar 2002 - Jan 2004, Vantaa, Finland
product
the operating application of an MRI scanner, where I worked on patient administration

After the dot-com crash I joined Philips. The company was cool and the product was even cooler. Developing an MRI system felt like helping to cure cancer. The console was a Windows application written in C++ and MFC.

Once I was told that I could not eat my Snickers bar at my desk, because we were developing medical equipment. Since then I know that certifications and regulations are not a joke.

What I learned

Regulated development teaches discipline. I also learned to do good work with tools I would not have picked myself.

C++ · MFC · Windows · medical device regulation

1998 - 2002 Nokia Networks · Software Developer (consultant) SMS and USSD centres. There is a good chance your text messages went through this code.
dates
Apr 1998 - Jan 2002, Helsinki
product
Nokia's SMS and USSD centres, one of the leading SMSC products of the time
team
100+ engineers

Headhunters recruited me from university straight to Helsinki. I developed the alarm and event logging daemons of the SMS and USSD centres. They were essential for operating the system and for producing the charging data records that billed the end users.

The control centre UI, used to read logs and alarms and watch the system daemons, was written in MetaCard. On my suggestion it was migrated to X/Motif on HP-UX.

USSD was fun to develop. It never really took off, even though it had more potential than SMS. The dot-com crash ended this chapter.

C · C++ · HP-UX · X/Motif · telecom messaging

1992 - 1993 Before the career · Kecskemét, Hungary First paid code at school, and a NetWare mail app with two friends.

1992, the first paid job. My uncle had written an accounting system in dBase/Clipper for a big Czech rubber company. Every function was started from the command line. I wrote a Turbo Vision user interface for it in a few weeks and got good pocket money. That was when I realised that programming can be a profession.

1992-93, "letter". In grammar school in Kecskemét, two friends and I wrote a mail application for Novell NetWare 3. We did not know the Internet existed yet. Each of us took different features every week to avoid conflicts, we talked every day, we exchanged code on floppies, and usually one of us merged. Without knowing it we were doing what I later recognised in Scrum. We even spent our pocket money on train tickets to ask the Novell importer why an interrupt did not return the full member list of a group. We were that dedicated and committed. That has not changed since, and neither has the passion to develop cool stuff.

Turbo Pascal · Turbo Vision · dBase/Clipper · NetWare · DOS interrupts

Secret weapons

Skills

Most CVs just list skills. Click a row to see what they mean in my case.

people People management Hiring, onboarding, one-to-ones, growing people, and the hard parts too.

Hiring

I document everything and take lots of notes. Hiring in Germany at SUSE, where documenting interviews is mandatory, taught me that. Good notes show me strengths and weak points long after the interview, or after dozens of interviews.

Referrals and my team's feedback weigh a lot, but never alone. I read CVs for substance rather than polish, and look for the right keywords and for signs of a real, original human.

A new team

I listen, ask questions and learn how they work. I look for the leaders and the followers and try to understand the social dynamics. I never suggest changes in the first few months.

One-to-ones

Weekly, with everybody who wants one. Some prefer to meet ad hoc, but most of my team mates like the weekly slot. It takes from 15 to 60 minutes. I always ask about success stories, problems I can help with and upcoming plans. I also ask how they are doing, and I listen.

Low performance

I approach low performance like a running coach and build on identity. Most engineers are proud of their craft, and that pride is where I start. Once the identity is back, the rest is habits, environment and workflow. It works most of the time.

A team mate lost momentum for a while. Bad habits stick just as strongly as good ones. In our one-to-ones we talked about problems they had solved and things they were proud of building. In parallel I helped them rebuild a routine, with short daily check-ins and pairing with team mates on tasks. After 2-3 months they were back.

Sometimes people do not get better. Then I look for the reason behind it. Low performance is often a sign of a bigger problem. If something personal is behind it, I make room for a break and point people to the support the company offers.

If nothing works, I ask HR to step in and we set up a formal performance improvement plan, following the legal procedure, and I see it through to its outcome.

Growing people

The same method. Build on their identity, improve their habits and help them build a name in the community. Some of my team mates were promoted.

A very good scrum master on my team wanted more technical work, but did not see how much it could do for him. I told him what I learned from my own career. Engineering skills degrade fast, and keeping your edge matters. He took on more technical tasks, and soon he was contributing code as well as running the team's process.

Conflicts

Most conflicts shrink when people feel heard. I ask questions first and listen.

Letting people go

I have let people go when company strategy changed or costs had to be cut. Not a fun job. Not every manager has done it, and I have.

projects Projects and delivery Plans, estimates, risk and pressure.

Estimates

Some engineers do not like to give estimates. I ask them what they would ask first if they hired a contractor to renovate their kitchen. Always the same two questions. How long will it take, and how much will it cost? The same applies to our work.

Risk

Skydiving taught me how to handle risk. Recognise the risks and follow the procedures that were developed to manage them, strictly. Risks do not go away, but seeing them, planning with them and following written procedures keeps them under control.

Pressure

My career is full of crazy deadlines. Working on a customer's live telecom system that makes millions of dollars a minute is pressure. Telling telecom executives how to run their network is pressure. Delivering a working UI Toolkit for a mobile platform on a hardware partner's schedule is pressure.

Coaching project managers

At Nokia I coached project managers and scrum masters to adopt agile methods, make realistic plans, build reasonable roadmaps and stay within budget. Coaching also meant listening to their problems and helping them convince the decision makers.

products Products and services Product owner at Nokia and Canonical, service business at Nokia Networks.

What not to build

Saying no is based on what the customer actually needs and what we have time and people for.

Services

A product is used, a service is received. A service sold to a customer must have an impact right away. Selling and delivering RRHC taught me to think about value from the first day.

Talking to executives

Find out what problems they have and how you can help solve them. Then link the technical solution to revenue or to cost reduction. That is the language they speak.

engineering Engineering An optimiser engineer. I do what nobody has time to do.

Either the dirty work, or the optimisation that people do not even notice. Faster tests, cleaner packages, tools that remove a manual step.

Languages and tools

My all-time favourite language is C. It is pure, clean and effective. For quick tools and quick fixes I use Python. I would not touch MFC or .NET again if it is up to me.

What I am proud of

Usually the most recent thing. I like the zypper changelog plugin a lot, because I made it for a real need. And from the old days, a dynamically interpreted, markup-based UI application on X/Motif, years before QML.

The home lab

Everything I play with at home ends up in my job. I experimented with local LLMs before we adopted them at SUSE. I first used Grafana in my Home Assistant setup, and later built Grafana dashboards for the SUSE maintenance update queue and for binary function coverage.

C · C++ · Python · Perl · Go · shell · Qt/QML · GNU/Linux · HP-UX · Solaris · BSD · RPM/OBS · Debian packaging · containers · cross builds

quality Quality engineering Quality is much more than testing.

Quality engineering starts with the product itself. Does it have good unit tests? Do its upstream tests run? What do competitors have that we do not? I compare our testing with Fedora, Debian and Ubuntu, and where they test a package better, we close the gap.

Automated tests are one tool among many. And if something matters, test it. I learned that from a blue button.

openQA · test automation · CI · binary coverage · upstream test suites

AI AI-assisted engineering How I actually use it, and what I never let it do.

I use opencode and switch between 4-5 models depending on the work. Expensive frontier models plan and design. Cheaper, simpler models follow the instructions the better models wrote. I think building a Karpathy-style LLM wiki for a team's knowledge is good practice.

I would never let an AI agent do system administration as root on a live system with business-critical load without a human watching.

opencode · frontier and local models · RAG · Ollama · ChromaDB

languages Hungarian, Finnish, English The same person in all three.

Hungarian is my native language and the one I read fine literature in. I recently read György Spiró's Padmaly, and Andrea Tompa is my personal favourite. I am better company in Hungarian when the topic is art, politics or social life.

Finnish is the language of my family and friends. It took me 6-8 years to become fluent, because Finns speak good English and politely switch to it when they see a foreigner struggling. I had to ask my friends to stop speaking English to me. Finnish is the most balanced of my three languages. At work, language is a tool, and I use whichever works best. Sometimes that is Finnish.

English has been my professional language for decades. Documents, code comments, manuals and meetings. I am better company in English when the topic is technology.

off work Running and skydiving Both taught me things I use at work every day.

Skydiving

1,800+ jumps, 500+ of them as a tandem jumpmaster. Skydiving taught me to respect procedures, standards and protocols, and what it means when somebody's life is in my hands. Every jump, the procedures saved my life. That is the idea.

Running

Marathons and half marathons, and once a 63 km ultra. Marathon in 3 h 15 min, half marathon in 1 h 30 min. Running taught me discipline, and to fall back on identity and habits when motivation runs out. Every race teaches me something, usually about myself. The more I run, the more I understand how I operate.

Studies

University of Szeged, Mathematics and Computer Science, 1995 - 1998