Skip to Content
How We Work

Collaboration and Customer Focus: Pure’s Machine Learning Team

At Pure Storage, every decision is in service of one goal: making customers’ lives easier. Creating the Machine Learning team was no exception.

“Pure had been collecting data from our storage arrays from day one,” explains Farhan Abrol, who joined the company as an engineer in 2014 and became head of machine learning for Pure1® three years later. “We knew at some point in the future, we’d be able to leverage that data to help our customers. We just needed to figure out how.”


In the more than five years since launching Pure’s first commercial product, the team had developed a deep understanding of its customers’ pain points. Most competitors offered complex storage tools with thousands of settings—and often, hours-long delays before data was delivered. But Pure was laser-focused on simplicity and approached storage as a customer product—and the customer was royalty.

Instead of forcing users to wade through dozens of performance metrics, for example, Pure developed a single, actionable number called “load,” which dramatically simplifies capacity management. As the team worked to further optimize events like upgrades and migration, they realized that machine learning (ML) would be key.

“We backed our way into ML,” Abrol says. “We’d zoned in on capacity and performance forecasting was the hardest problem facing our customers. Then it was, ‘Okay, how do we solve it?’ And we realized we had this data.”


End-to-end Ownership

After launching its first project—Workload Planner, a “crystal ball” based on the AI-driven Pure1 Meta® platform—the ML team continued to grow. At the time, data scientists at most companies were working on consumer products, the opportunity at Pure was different.

“Because we’re B2B, every customer interaction is important. We can’t just write off a specific occurrence as an outlier,” explains Abrol. “We think about the output of our models in terms of absolutes, not averages.”

To stay close to customers, the team collaborates regularly with Pure’s UX designers and product managers—and they talk directly with users. “We attend Pure//Accelerate, Pure’s annual conference so we can see firsthand how customers are responding to the features we build,” says Abrol.

He says his team’s unique level of visibility into the customer experience has also created a unique culture of ownership. “A lot of companies have separate teams for assets, data planning, model research, and deployment,” he explains. “But because we understand the entire user story, we’re able to own things end-to-end.”


Humility and Teamwork

Another key cultural value across Pure is humility, Abrol says. “We’re honest about mistakes here. If you try something new and it doesn’t work, you’re not going to be penalized for it. The goal is to distill learnings the company can use.” For the ML team, each project offers new opportunities to learn and to collaborate with other Pure team members. For example, timestamps were a challenge at first. “We had one way of defining past and future, and the front-end team had another,” Abrol explains. “We got on a call together and figured out what it should look like for the customer.”

Data quality issues also provided an early lesson. Alerted to some problems by customer feedback, the ML team started by asking the data generation team to weigh in with initial theories. Then they analyzed and segmented customer data to identify specific subsets their colleagues could study further. “We built out an iterative pipeline,” Abrol says. We’d surface issues, they’d identify improvements, and we’d rinse and repeat. We were able to bridge the gap between our specialists and what customers were seeing.”


No false boundaries

In the months ahead, the ML team will expand beyond optimization and plans to dig into a new area: discovery. As always, simplicity is the number one goal. 

“We’re measuring storage, virtual machine analytics—all these interdependent metrics that can help customers identify problems in their businesses,” Abrol explains. “But then finding the source of those issues can be like looking for a needle in a haystack. We’re figuring out how to give customers more guidance, so they know where to focus right away and can start working on a solution.”

In the long term, Abrol says, the next big project could come from—and be led by—any member of the team. “We aren’t driven by seniority. If you have a good idea and feel confident you can run it, that’s what matters.” 

As for what that project will be? The guiding light will remain the same: Whatever’s best for the customer. “There are no artificial boundaries,” says Abrol. “When you see something that needs to be done, you talk to who you need to talk to and build what you need to build. That’s what Pure is all about.”


You’d Look Good in Orange

Don’t sit back and watch the world go by. Be what the world is watching.

Your Browser Is No Longer Supported!

Older browsers often represent security risks. In order to deliver the best possible experience when using our site, please update to any of these latest browsers.