Infrastructure UI

In 2014 Nutanix launched Prism (their infrastructure management platform) as part of their NOS 3.5 release to replace a flash enterprise interface with a browser-native console. I joined to make datacenter complexity feel as clear as consumer software. Later the challenge as we grew became to keep that clarity as Nutanix added clusters, products, and features in the years to follow.

The problem

Before Prism, IT administrators jumped between hypervisor clients, storage tools, and network panels to answer one question: is my infrastructure healthy? Simple tasks took multiple screens. Remote access often meant using Flash or Java plugins.

Nutanix had converged compute, memory, and storage in the stack. The interface had to show that convergence, not hide it behind separate tools. And it had to stay legible as the platform grew from a single cluster to multi-site operations, disaster recovery, lifecycle management, and dozens of new entity types.

My role

During the founding product years at Nutanix. Prism Element was our cluster console. Prism Central followed with NOS 4.0 as Nutanix's answer to management at scale: one HTML5 interface across datacenters.

I owned product design across lifecycle management, foundation imaging, storage management, dashboards, navigation, and alert patterns, while setting visual direction for the product. Over time that scope grew into design system leadership and visual design for Nutanix product surfaces company-wide.

By 2026 I had designed more than 32 features inside Prism, led helped formalize the design system and visual design teams as their manager, and watched the organization grow from four founding designers to 72+ across global sites.

The adoption program, React library, and accessibility standards from that era are documented in the Design System & Accessibility case study.

Process & decision rationale

Prism was never a one-time UI refresh. These decisions carried from the first HTML5 screens through twelve years of platform growth:

• Health before depth. Administrators needed cluster status in seconds. We led with scan-first dashboards and severity-coded alerts, and pushed advanced configuration off the default path. Less discoverability for power features, more confidence during incidents.
• One console for the full stack. Instead of mirroring siloed hypervisor tools, we mapped compute, storage, and network to shared objects so one navigation model covered the hyper converged stack.
• HTML5 and API parity together. Browser-native delivery mattered because plugins blocked SSO and mobile workflows.
• Navigation that could survive scale. We paired entity-based side navigation with search early, knowing Prism Central would eventually host more features than any menu could list.
• From patterns to enforcement. By 2024 guidelines alone were not enough. I led the shift to measured adoption: audits, shared tooling, workshops, and release-by-release targets until library usage reached 100% across the design organization. That work is the focus of the Design System & Accessibility case study.

Trend and capacity charts had to read at a glance on busy operations dashboards, dense enough for production data, not demo-friendly sample strings.

Outcomes & evidence

Prism Element became the unified home for health, VMs, storage, hardware, and alerts. Prism Central extended that language to multi-cluster operations: health, usage, and upgrades across datacenters from one interface.

Nutanix was among the first enterprise infrastructure vendors to ship a management UI in HTML5, and Prism remains the gateway to the platform today.

The work compounded. Design library adoption moved from 43% to 100% between 2020 and 2023, with release-by-release metrics, Level Access audits, and a dedicated design system team.

Founding side navigation and dashboard patterns, still the backbone of how operators scan infrastructure at scale.

Reflection

Enterprise admins do not want delight. They want confidence.

The hardest calls were subtractive: designing for long VM names and verbose alert text, accepting information density consumer UX would reject, and hiding power until context demanded it.

The longest bet was organizational. Components were the easy part. Keeping dozens of designers aligned across twelve years meant proving progress with numbers, validating patterns against real edge cases, and giving teams a system they could trust more than their own one-offs. That is what turned a founding console into a platform habit.