Founder’s Story: Qevlar AI
A chance connection in a college corridor brought Ahmed Achchak and Hamza Sayah together to create Qevlar AI
University is well known as the foundation for many lifelong friendships, and in the case of Ahmed Achchak and Hamza Sayah, this was no different. Hamza was studying a Bachelor’s degree in applied mathematics at prestigious Swiss university EPFL; Ahmed was doing his MSc in biotechnology and mathematics at CentraleSupélec, one of France’s great ecoles. Both universities offer a joint degree programme for a handful of their most talented students — and it was when Hamza undertook the joint degree program that he met Ahmed.
And so by chance, in 2015, the two found themselves as both coursemates and neighbours in their student accommodation. “His room was exactly next to mine,”says Hamza.
The two students found much in common and ended up collaborating at a Huawei hackathon held in Paris in 2018. The goal was to design an AI-powered anomaly detection tool for kernel-related cloud logs using real-world Huawei cloud data. . “We did pretty well at the hackathon,” says Ahmed. In reality, they won first prize — and learned that the tech they created ended up beating some of Huawei’s own benchmarks.
The vignette made both realise they had a shared talent — and a passion. “We grew convinced that we wanted to stay in the AI and cybersecurity field,” says Ahmed. “That’s what gave us the idea for Qevlar AI.”
But while the spark of interest in cybersecurity was lit in both Ahmed and Hamza, it would take several more years from that 2018 hackathon for them to establish the company.
One reason both decided against jumping into building their own company was a comparative lack of experience. “I really wanted to learn entrepreneurship first, because I came from an engineering school,” says Hamza. Another reason, says Ahmed, was that the tool they developed at the Huawei hackathon was only part of the idea that would eventually become Qevlar AI. “It simply wasn’t the right timing,” he says. “What we did for Huawei was basically a detection AI, and you had a lot of new anti-viruses doing the same thing. There was no real breakthrough.”
But the generative AI revolution inspired an idea that would prove to be a breakthrough — one that would propel Qevlar AI into the forefront of businesses’ minds. Hamza, who had risen to become head of research at unit tests generation platform Ponicode, had long been tracking the development of what we now know as large language models. At the same time, after a stint as chief technology officer at second-hand auto parts reseller Minautor, Ahmed had grown in confidence and felt ready to start something of his own.
The two friends reconnected and thrashed out the idea for Qevlar AI — an autonomous AI system that can automate the roles and tasks ordinarily done by organisations’ security operations centres (SOCs). But fine-tuning the product necessitated research, and a deeper understanding of the challenges practitioners faced, so they could truly serve them. Qevlar AI’s systems can interpret technical data organisations retrieve through their investigations, then extract insights to drive investigation orchestration and influence outcomes, as well as summarising information in a digestible format for analysts to act upon.
“We reached out to anyone who wanted to talk to us,” he says. “We literally sent hundreds and hundreds of messages to people on LinkedIn, some of whom we knew, some of whom we’ve never spoken with before.” Respondents were quizzed about their pain points when it comes to operating their SOC.
One of the main issues raised was that staff working within an organisation’s SOC were exhausted from switching between mundane repetitive tasks, then having to leap into action when something serious went wrong.
“People were flooded by alerts to the point that they would get burned out,” says Ahmed. “That’s not sustainable, and something we definitely wanted to solve.” But the conversations also made them realise that SOCs can be a source of insights for how the wider company is operating. “Because the AI is able to discover what things are working or not working in your detection systems, it can actually be a positive feedback loop for the organisation,” he says.
The simple but powerful choice Qevlar AI made was to move SOC’s from an alert-centric to an asset-centric model. “The idea is to come back with contextualised insights that help steer the defence posture of the organisation,” says Ahmed.
Starting to work with design partners in March 2023, Qevlar AI began refining their product based on real-world feedback. The official unveiling came with the announcement of their seed round, drawing significant industry attention and momentum. Qevlar AI would supercharge organisations’ security operations centres by prioritising and remediating risks and threats through autonomously investigating issues and raising alerts. The duo began showing Qevlar AI to CISOs of large organisations around Europe, and were blown away by the response.
Ahmed recalls many CISO’s had never seen anything like it and were very interested in trying it out, which gave them confidence to keep building and iterating..
“By having conversations with people from SOCs, we built a small ecosystem of professionals around us, giving us feedback,” says Hamza. That early feedback was vital, adds Ahmed. “We’re not from the field ourselves, so we have to be very disciplined about this feedback loop that we have,” he says. “That enabled us to continue refining our hypothesis on what people in the field want.”
That ecosystem included Mehdi Ghissassi, Director & Head of Product at Google DeepMind, who recommended a meeting with them to Julien Hobeika, partner at EQT Ventures.
“What struck me when I met them was the level of energy they had,” says Julien. “Just in the code, the level of patience, and the passion.” The meeting between EQT Ventures and Qevlar AI was meant to last 30 minutes, but Julien extended it in order to see a terminal-based demo of the product. “I think they saw a lot of interest in the market for the solution,” says Ahmed. “The way we were framing things and the approach that we wanted to have made sense to him. And I guess that’s what contributed to moving forward with us.”
Things moved quickly from there: “It took less than 10 days,” says Julien, for EQT Ventures to lead a $5 million funding round. “When we have high conviction, we move very fast.”
The investment and the support helped the founding team to grow Qevlar AI. “EQT Ventures turbocharged our hiring strategy and process. They linked us with major commercial partners who became clients and ecosystem allies offering both insights and hands-on support. Their constant guidance and role as a sounding board have been game-changing for us,” says Hamza. That support — both financial and advisory — has allowed the two-person team to grow to an 11-strong business, with plans for further hiring.
And that’s just the beginning. “My vision for them is huge,” says Julien. Already, the demand for the technology is widespread within the industry. “Today, we have companies that are using our technology in production in France, Germany and the Middle East,” says Ahmed. “We have a bunch of large international organisations that are testing our technologies in production.” The service is used by major companies across Europe, while the business is part of an AI program powered by Meta and Hugging Face at Station F, and can reduce the average response time to critical alerts by 30%.
Those companies have bought into Qevlar AI’s promise, as outlined by Hamza. “I see Qevlar AI is poised to be the game-changer in cybersecurity, eliminating SOC inefficiencies and fortifying organisational defences like never before.” ,
