Learning that the customer's project is the real production environment

I grew up in Jonquière, in the Saguenay, where my father was an electrician at an aluminium smelter. He used to say that a potline is never switched off on purpose, because if the metal freezes in the pots the restart takes months. I did not understand then that he was describing a design principle. I understand it now: some systems have to be changed while they run, and the people who change them must know exactly what depends on them.

I studied computer engineering at the Université de Montclerc, graduating in 1999, and stayed for a master's degree in computer science, finishing in 2001 with a thesis on rendering large terrains at a steady frame rate. That work took me to Cordillon Simulation in Montréal, which built the visual systems for airline flight simulators. In my second winter there, a change I had made to terrain detail passed every test in our laboratory and then made the image stutter on one customer's simulator during a night approach in falling snow. Instructors stopped using the device for two days. Nobody was hurt and the fix was small, but I learnt that a test environment describes our assumptions, not the customer's conditions.

Games drew me because rendering was the product rather than one component of it. In 2004 I joined Studio Rivenord, an independent console developer in Québec City, as a senior rendering programmer, and led its rendering team from 2006. We shipped two console titles on an engine we had written ourselves, and I learnt how much of a studio's schedule is decided by tools nobody outside the studio sees. In 2008 I moved to Grisfell Games Montréal, a studio of 420 people owned by Valdrecht Entertainment, a fictional Swedish publisher, as lead engine programmer, and became its technical director in 2010. Grisfell shipped a large open-world game on its own engine in 2011. The engine was impressive; the nightly build took nine hours, and I spent a year getting it to two.

In 2012 Valdrecht's first online game launched badly, and the publisher asked me to build an online engineering group in Montréal that would serve all its studios. I had never run a production service. For four and a half years I learnt how, mostly in the early hours of Sunday mornings. We built a shared platform for accounts, matchmaking and game servers that carried nine titles and a peak of 1.9 million concurrent players, and we raised availability from about 99.5 per cent to 99.93 per cent. The lesson I took from it was cultural rather than technical. Outages became rarer when engineers could describe their own mistakes in a review without fearing for their jobs, and when product managers saw the cost of each incident in the same report as the features they had asked for.

By 2016 I wanted to build online services as a product that studios would choose, rather than a cost centre that one publisher's studios were obliged to use. Sternhollow, then a Vancouver engine company of about 700 people, offered me the chance to turn a beta into a business. I moved my family west in January 2017. As vice-president of engineering for online services I grew the team from 30 to 150 people and the service from nine titles to 140, with revenue rising from C$19 million to C$71 million by 2020. When the company's first CTO, one of its three founders, stepped back to an advisory role, our chief executive asked me to succeed him in March 2021.

As CTO I am responsible for the engine, the editor and tools, Sternhollow Online, our infrastructure, security engineering and developer productivity. I have seven direct reports, present to our board each quarter and report to its Audit and Risk Committee on security and reliability. The decisions that have defined my tenure all involve customers we cannot see: disclosing a licence problem in code that shipped inside 412 games, freezing feature work after an outage, buying GPU servers instead of renting every hour, and refusing a clean break in the engine's core while studios were mid-production.

I lead through written reasoning. Any significant technical decision at Sternhollow starts as a short paper that states the alternatives, who disagrees and what we would give up, and the people who disagree are invited to write their own section. It slows some decisions down by a week. It has also meant that when we have been wrong, we could see where.

I am probably known in the industry for three opinions: that an engine vendor's build system is as much a production system as its servers, that a reliability problem is usually a product-management problem first, and that AI coding tools help engineers a little and reviewers not at all unless someone plans for it. Each was learnt by paying for the alternative. Alongside my executive role I serve on the board of the Westerly Game Makers' Society, a non-profit that supports independent developers in British Columbia, and I am preparing for one independent directorship outside games.

Open to Board appointments · Speaking invitations

Languages / French · English