How Flux built a GPU-powered AI Infrastructure 90 Days ahead of schedule

The Challenge 

The field of AI is possibly the fastest-moving sector in all of history. Almost every day, new developments tilt the playing field, causing players and goalposts to slide around. These shifts aren’t just happening within algorithms or applications; the very foundation of computing is changing.

Startup Flux wanted its decentralized network to weather the AI field’s rapid shifts. Already aware that GPUs are invaluable for AI’s parallel processing demands, they embarked on an ambitious project to build decentralized cloud infrastructure for AI and general compute services via enhanced GPU connectivity.

However, the Flux team was extremely short on one very critical resource: time. To stay competitive in the lightning-fast market, they needed to establish their platform as quickly as possible: “Every day that we lose is a day that someone else can beat us to the punch,” said Davy Wittock, Chief Business Officer at Flux. 

As a self-funded startup without VC backing, Flux was strapped for adequate resources to afford a Kubernetes expert. Working in Kubernetes requires a deep understanding of container technology, microservices architecture, and large-scale system administration. Experts have been in high demand ever since Kubernetes became the de facto standard for deploying and scaling cloud applications.

“Those folks are needed, and sought after, and usually quite expensive,” said Wittock. “To find someone to do [Kubernetes], and do it part-time, seemed almost impossible.” 

“We did the traditional recruiting stuff: LinkedIn, Monster, reaching out to people we met at conferences. But we found ourselves in that old-school lengthy recruitment process: interviewing, going through resumes, setting up calls. We had no time for that,” Wittock said.   

Flux turned to A.Team for help, leaning heavily on AI-powered search to find a highly-specialized builder. They weren’t just looking for a general Kubernetes expert. They wanted someone with deep Kubernetes experience, who could work flexibly and advise on working in brand new markets. 

Ideally, Flux’s A.Teamer would have experience building category-defining products, because that’s what they were doing. Wittock said, “People need to understand: the skillset we were looking for is so specific. But within a couple days, we had the right fit. A week later, he was already working with the team.”

A.Team’s AI-powered platform took Flux’s niche requirements and identified the right talent from A.Team’s network of 11, 000+ product builders based on their work experience, past project data, and working relationships: “A.Team’s AI was a big recruiting shift for us. We basically see it as the next evolutionary step in recruitment,” Wittock said. “If you ever post a job on LinkedIn, the amount of responses you get is ludicrous. We had people spending weeks trying to filter through them. But the AI that A.Team created has really sped things up, from weeks to minutes for our recruitment.”

{{quote}}

The Build 

While other candidates had Kubernetes experience and classic technical skills, the A.Team builder they matched with, Nick, had a background in AI, cluster building, and new tech services. He could understand the specific demands around decentralized computing shifts, AI workloads, and optimizing GPU clusters for machine learning tasks. And most importantly, he had built with them before. 

“Nick brought in so much expertise to the team,” said Wittock. “We were looking for an out-of-the-box thinker with experience in on-demand compute and decentralized data centers, with the skillset to think on their feet. Nick was a perfect fit.” 

Nick “hit the ground running” and immediately helped to propel the project forward. According to Wittock, “Without Nick, this project would be delayed and I’m not sure we’d even have a release by now.” 

Nick came to the team with ideas, and implemented a system wherein the team could “punt ideas around and compare pros and cons.” That was the benefit of having an experienced hire who had worked in the field. “Having someone who was already aware of some of the pitfalls inherent to Kubernetes really helped us move quickly along,” said Wittock. 

With Nick on the team, Flux was able to release their decentralized AI platform 90 days ahead of the projected deadline.

{{quote-2}}

Outcomes

The Flux team is now officially ahead of schedule. They’re incorporating B2B models that were initially planned for the end of the year. “In many ways, we’ve actually exceeded our deadlines,” said Wittock. 

Flux’s team has also grown more independent around resolving technical issues with the help of their A.Team: “Nick hasn’t just augmented the project with his individual expertise. He’s taught the rest of the team what to do when an issue happens and he’s not available,” said Wittock. 

Flux can now onboard customers that require the largest graphics cards in existence, with GPUs ranging into the thousands. “We’re not afraid of tackling these types of contracts anymore, so we’re well ahead of the curve when it comes to our project,” said Wittock.

Flux now has tens of  thousands of GPUs registered on their network. Continuing to grow at this scale “will allow us to work at enterprise-level with the top 500 companies,” said Wittock. And users are “raving about the ease of use and how easy it is to deploy something.” 

Nick was instrumental in ensuring that for Flux, user data had adequate safety and security controls in place. “He has basically created an easy-to-use software that every day consumers can benefit from without having the expertise required at the back-end,” said Wittock. 

The industry is already paying attention to what Flux is building. According to Wittock, “We already have big AI players talking with us because of software we created that’s still in alpha. The impact we’re creating in this space thanks to A.Team is seen and felt in the world.”