Dmitry Fonarev

Ask most engineers what they think about when they think about software testing, and you will get a sigh. Fonarev built a company on that sigh. Testkube, which he co-founded and now runs as CEO, takes the messy pile of test tools every team already owns - API checks, load tests, end-to-end suites, security scans - and orchestrates them directly inside Kubernetes. No bespoke glue. No CI/CD pipeline groaning under the weight of a thousand parallel tests.

The pitch is unfashionably specific: decouple testing from the pipeline, run it where the application already lives, and let it scale the way the cluster scales. By September 2025 the platform had powered more than 100 million automated tests. The logos behind those numbers are not toys - Cloudflare, Siemens, Adobe, Volvo.

Then came the money. In September 2025 Testkube announced an $8 million Series A led by Ratmir Timashev, the co-founder of Veeam, alongside Insight Partners. For a testing tool, that is a loud vote of confidence. Mike Triplett of Insight put the thesis plainly: testing is "the next big bottleneck," and the bottleneck is getting worse as AI writes more of the code.

That sentence is the whole company in a breath. When code is generated faster than any human can read it, the old rhythm of write-then-test falls apart. His co-founder Ole Lensmar - the engineer who created SoapUI, one of the most downloaded API testing tools ever made - calls the platform a "safety net" for a moment when "decisions are happening faster than humans can follow." Fonarev is the operator who turns that conviction into a product roadmap, a customer list, and a payroll.

How a side project became a company

Testkube did not arrive fully formed. It started as an open-source project living inside Kubeshop, the developer-tools studio Fonarev co-founded with Lensmar in 2021. The idea was a workshop for next-generation tools aimed at developers, testers and DevOps engineers. One of those experiments - a way to run tests natively in Kubernetes - kept pulling ahead of the others. In 2023 it spun out as its own company. By October 2024, Fonarev had moved from Kubeshop's CEO chair into Testkube's.

It is a familiar arc told in reverse. Most founders look for a problem to chase. Fonarev had spent two decades inside the problem and only then built the company to solve it.

Two decades on the quality side of the house

Before he was a founder, Fonarev was the person large companies trusted with the unglamorous machinery: engineering, quality, and the customer support lines that ring when quality slips. He earned a computer science and mathematics degree at Boston University and started out writing code at Nortel - close enough to the metal to know what breaks and why.

From there the resume reads like a tour of software's quality frontier. Vice President of Quality at IntraLinks. Vice President of Software Development, QA and Customer Support at vKernel from 2011 to 2013. A turn as Vice President of Engineering at Dell. Then, from 2014 to 2020, Executive Vice President of Engineering and Customer Support at SmartBear Software - a company whose entire business is testing and quality tooling. He did not stumble into testing. He spent a career there.

There is a pattern in that list worth noticing. Engineering and customer support keep appearing together in his titles. That is not an accident of org charts. The people who run support hear the consequences of bad quality first, in real customers' voices. Spend years on both sides of that wall and you stop seeing testing as a chore. You start seeing it as the thing standing between a company and its angriest emails.

The unfashionable bet

Plenty of startups in 2025 reach for the brightest object in the room. Fonarev's bet is duller and, if he is right, more durable: that as software gets faster and more autonomous, the boring discipline of testing stops being optional and becomes the constraint on everything else. The teams shipping AI-generated code the fastest will be the ones who can verify it the fastest. Testkube wants to be that verification layer - sitting inside the cluster, scaling with it, asking the unglamorous question over and over. Does this actually work?

It is a question Fonarev has been asking in one form or another since Nortel. The tools changed. The clusters arrived. The code learned to write itself. The question never moved.