Pure Storage's 2,000-engineer team is building a new development methodology powered by Augment's Context Engine—aiming to triple their delivery velocity while maintaining enterprise-grade security.
The challenge: Scaling AI adoption across a massive codebase
When Dr. Ratinder Ahuja, CTO, Security & GenAI at Pure Storage, first presented his vision for a "GenAI-transformed enterprise" to the board, he knew the biggest opportunity and challenge would be transforming how 2,000 engineers work.
Pure Storage isn't your typical software company. As one of the world's leading data storage providers, they operate with an enormous, interconnected codebase. "Think of us as an embedded system—everything connects to everything," Ratinder explains. "One small change in a device driver can have massive implications across other parts of the system."
"We want this engineering force—product management, support, and other technical functions—to all move at a certain velocity," Ratinder explains. But when you have thousands of engineers, each with their own approach, their own habits, and their own level of expertise with new AI tools, you can't realize the benefits of these systems.
Their engineering team had already experimented with GitHub Copilot when Microsoft made it available. But in a monolithic repository with C++, wrappers, and multiple languages all intertwined, traditional code completion tools struggled.
With our huge monolithic codebase, we found that Copilot wasn't really adding value Dilip Kumar Uppugandla, VP of Engineering
The team needed something different—a tool that could understand their complex, interconnected systems and deliver consistent results across thousands of engineers.
Why Augment: Context and security at enterprise scale
When Pure Storage evaluated AI coding tools, two dimensions mattered most.
First, security. With a secure software development lifecycle at their core, they couldn't compromise. "Augment was the only company taking enterprise security policies seriously," Dilip notes. For Ratinder, whose dual role encompasses both security and GenAI, this was non-negotiable. Augment enables teams to embed security requirements directly into code generation—following secure development processes, identifying vulnerabilities, and suggesting patches that would otherwise never be found.
Second, context. Pure's massive, interconnected codebase spans multiple product lines, languages, and teams—with layers of technical debt.
"Augment's Context Engine was able to navigate all of that," Ratinder explains. "Other tools couldn't understand when a function is wrapped around a C++ wrapper that goes all the way down to an actual device driver call. Augment could."
They selected a cross-organizational team to evaluate tools. Augment won.
From skepticism to "over my dead body"
Adoption wasn't immediate. "Initial results were positive, but it took time," Dilip remembers. "Once engineers tried something and didn't have a good experience, it was very hard to get them to go back to any tool."
But something shifted when teams discovered Augment's agent capabilities and CLI. Engineers who truly learned to leverage the Context Engine became passionate advocates.
“Engineers told us: 'You will not take this away from us. Over my dead body.' There's so much passion because with a complex codebase and a big monolithic repo, Augment solves problems that nothing else can.”
The breakthrough moment came when Ratinder worked on a hobby project to stay current with the technology. He defined the problem mathematically—state machines, truth tables, all the classical techniques.
"The system ran 1,000+ tests completely—everything worked—and I never even looked at the code," he recalls. "That was transformational. You can trust the system because you guided its construction through formal specifications."
That's when the vision crystallized: What if all 2,000 engineers could work this way?
Building the methodology: From stars to standardization
The goal at Pure Storage isn't just to have individual stars, but to establish a consistent baseline across the entire organization.
"Our first goal is to establish a consistent baseline—how you work with Augment, how you define what we call 'recipes' or instructions," Ratinder explains. "That way we ensure 2,000 engineers are all running at the same pace, and nobody gets left behind."
Stage 1: Formal requirements and design
Business requirements, marketing requirements documents (MRDs), product requirement documents (PRDs), engineering response documents, technical design documents—this volume of work historically got short-circuited because it took too much time.
"The last 15 years these processes have been short-circuited because it takes time to build them," Ratinder notes. "But Augment can actually build those with automation as well. Now the subject matter experts can create these formal documents and review them. You have a GenAI assistant in building and reviewing them, so the human expert still looks at it, but it's so much easier to digest complex information."
“Security requirements have always been a pain in the neck—you've got to go find a security guy to tell you what the security requirements are. Now that can be automated with things like Augment.”
The key is standardization: "How the spec is created" and integrating consistent terms across the organization. When everyone uses the same formal methodology for defining problems, AI can work consistently across all teams.
Stage 2: Mathematically complete design
Not every engineer knows formal methodologies—truth tables, state machines, formal messaging protocols. But these techniques are exactly what enable reliable AI-generated code.
"Many engineers may not be trained in formal methodologies," Ratinder acknowledges. "They may not know how to do truth tables and state machines and formal messaging protocols and things like that. But GenAI can assist, and you can actually build a formal, provable design—mathematically complete design, no livelocks, no deadlocks, no runaway processes."
This is the breakthrough: AI doesn't just write code. When given formal specifications, it constructs systems that are provably correct.
Stage 3: Security embedded at every stage
Security isn't an afterthought—it's woven into every stage of the methodology. From threat modeling to code scanning to automated generation of least-privileged security policies, security requirements that used to require specialized expertise can now be systematically applied.
"Our software development lifecycle is actually called the secure software development lifecycle, and it is rooted in a practice called DevSecOps," Ratinder explains. "With Augment, you're able to provide those instructions and say, 'Follow the secure development process and ensure that the code that is generated does not have these vulnerabilities."
Augment can also "provide more context on existing vulnerabilities and have it go find those in the code, come up with solutions to patch those, or address those. For security, it's such a great service to have something like this that can address issues that would never be found or addressed otherwise."
Stage 4: Formal testing with comprehensive coverage
With specifications clearly defined and designs mathematically complete, comprehensive test suites follow naturally—accelerating quality assurance while maintaining accuracy.
"Once you have those formal specifications and designs, then the formal testing can take over," Ratinder says. "You can see how every stage, including security embedded in all these stages, gets accelerated. That's what we're trying to standardize on, so that the coding agents can deliver this at a higher velocity than ever possible before—and accurately. That's a really important part. It needs to be predictable and accurate."
Pure Storage's AI-powered development methodology
The audacious goal: 3x velocity
Pure Storage isn't being modest about their ambitions. "Our overarching goal for the next two years is to triple the velocity of delivering outcomes," Ratinder states. "For instance, if we're shipping one feature a year, we want to ship three."
This isn't just about writing code faster—it's about fundamentally transforming how engineering work gets done. Beyond core product development, teams are finding innovative applications across disciplines: automating secure AWS deployments, building bots for finance departments that connect Salesforce, NetSuite, and SAP, and enabling non-technical staff to become "citizen developers."
"We believe that as we use more and more Augment, and as we get 2,000 engineers all trained on the methodology, 3x velocity is achievable," Ratinder says confidently. The measurement is clear: 3x velocity, with predictable, quality, secure outcomes.
What's next: AI as tutor and methodology guide
As Pure Storage continues scaling AI adoption, they see Augment evolving into something more than a coding tool—it becomes a tutor and methodology guide.
"Augment sees everything that's happening across the organization, and it can surface best practices—saying, 'For this class of problem, here's how you should approach it,'" Ratinder envisions. The tool itself can recognize patterns across the organization and recommend best practices, helping every engineer—not just the stars—work at the highest level.
For a company serving some of the world's largest organizations with critical data storage needs, the stakes are high. But as Dilip puts it: "We're definitely happy so far with what we see with Augment, and we're confident we can continue to have a strong partnership and get even more results."