// 01
The client
A Gothenburg-based technical recruitment firm that screened and placed software engineers for companies across the Nordics and beyond. When our engagement began in 2014, their candidate-testing tool was a WordPress site running multiple-choice quizzes. They had a bigger vision: a real platform that could test how well someone actually codes — not just what they remember.
We worked together for seven years turning that vision into one of the more sophisticated technical-hiring platforms in the region.
// 02
The challenge
Testing knowledge with multiple-choice questions is easy. Testing engineering ability is hard — and doing it at scale, fairly, and without the system being gamed or broken, is harder still.
Three problems stood out:
1. Running untrusted code safely. To test real coding skill, you have to let candidates write and run actual code — in whatever language they choose. But running strangers' code on your servers is a security nightmare. One malicious or runaway submission could compromise or crash the system.
2. Testing skill, not just correctness. Two candidates can both solve a problem. One is good; one is exceptional. A binary pass/fail can't tell them apart — and telling them apart is exactly what recruiters are paying for.
3. Surviving a hackathon. Testing one candidate is one thing. Hosting a live hackathon where ~1,000 programmers hit the system simultaneously is a completely different load problem.
// 03
What we built
Over the engagement, the platform evolved from a WordPress quiz tool into a full applicant tracking system (ATS) and technical assessment engine — rebuilt as a Node.js + React SaaS. The pieces that mattered most:
A secure code-execution sandbox. When a candidate runs code, we spin up a tiny, isolated AWS instance loaded with just the compiler and libraries needed for that language — run the submission against known test cases (visible to the candidate), hidden test cases, stress tests with very large inputs, and edge cases — capture the output or errors, and tear the instance down the moment it's finished. The engineering tension was constant: isolation and safety pulled one way, cost and speed pulled the other. Balancing them — fast enough for a live test, cheap enough to run at scale, safe enough to never be exploited — was the hardest problem in the build. By the time our team last worked on it, the sandbox supported 11 programming languages, from C++ and C# to Python, Java, PHP, and JavaScript.
Adaptive testing with chess-style ratings. This was the feature recruiters loved most. The founder, a strong chess player, suggested we borrow the idea behind chess rankings — Elo ratings — and apply it to test questions. Each question carries a rating; each candidate effectively plays against the questions. As a candidate answers, we update the question's rating and decide, in real time, what to serve them next. The result is a test that adapts to the person taking it — quickly separating genuinely gifted engineers from merely good ones. That distinction is the whole point of technical hiring, and the adaptive engine made it visible.
An ATS built like a Kanban board. The hiring workflow was visualized as a board — candidates moving through stages — with automations, an internal messaging system, and multi-recruiter support. Companies were organized into departments, each with multiple recruiters, posting jobs and managing applicants through the pipeline.
Hackathon and challenge hosting. Beyond one-at-a-time testing, the platform could host live coding challenges and hackathons — which is where the real load-testing happened: keeping the system responsive while ~1,000 programmers competed at once.
As the platform grew, it became valuable enough that it was eventually spun out as its own company — separate from the recruitment firm — to avoid the conflict of interest of a recruiter also selling the tool to rival recruiters.
// 04
My role
I started as the lead developer — hands-on from the original WordPress platform through the first SaaS version (Node.js with Angular, later moving the frontend to React). As the engagement grew, I moved into delivery management — scrum master for our team, and the partner who sat in the client's strategy and roadmap calls. At its peak, our Gyrix team of five (developers, leads, QA) worked on the platform.
It's the arc I still work in today: deep enough in the code to build the hard parts myself, senior enough to own the relationship and the roadmap.
// 05
The outcome
- →
A seven-year engagement (2014–2021) — the kind of longevity that only comes from consistently delivering.
- →
Live hackathons with ~1,000 concurrent candidates — load-tested at real scale.
- →
The platform recruiters relied on to place candidates at Spotify, Boeing, and other major companies — with individual hiring managers at those companies using it directly for their own teams.