OpenShift Data Science, Tech Preview
Red Hat's first foray into AI/ML platforms started as an open-source project with no UI and no clear audience. I defined the users, mapped the journey, and shaped the product's first shipped experience.
01 · Context
An open-source project with no UI, entering a market Red Hat hadn't touched
OpenShift Data Science began as Open Data Hub, an open-source project meant to simplify the sprawl of tools available for building and deploying AI/ML models on Kubernetes. Red Hat had deep credibility in open source, but none yet in data science or AI/ML, this product was that first step.
02 · Understanding the product space
No UI meant starting with the users, not the screens
When I joined, the product had no UI yet. My first job was figuring out who it was for. I learned the fundamentals of data science and the tools data scientists rely on, and ran countless informal conversations with product, engineering, and prospective users to gather their needs and plans.
I ran a workshop to align the highest-level stakeholders on a shared problem statement, our users, and where the product needed to go.
From there I mapped the high-level needs of each persona and, working weekly with stakeholders, built a journey map showing which users owned which part of the process. Those weekly sessions became the team's core touchpoint, the one place everyone discussed problem-solving and the end-to-end experience as we got closer to a Tech Preview release at Red Hat Summit.
03 · First concepts
Designing the highest-priority moments first
With needs and journeys mapped, I started on the pages the team agreed mattered most. An installation pattern already existed, so I explored augmenting it — letting installers choose what end users would see and how the product would behave.
Inside the product, we knew users needed one place to reach the product's tools: learning content, a way to launch Jupyter notebooks, and a status dashboard for both users and admins.
04 · Re-scoping the MVP
Cutting scope to hit a small team's GA date
Once we dug into the technical lift for the augmented install flow and notebooks, it was clear the work exceeded what our small team could deliver by the planned GA date. We refocused the Tech Preview and GA on the application pages and learning content, and leaned on the existing installation pattern instead.
05 · Final pages
Building on PatternFly for speed and consistency
Using PatternFly, Red Hat's design system, I built pages from existing components so the options stayed easy to implement and understand. A card view quickly emerged as the clearest way to present applications, and we carried that same pattern into the documentation page for consistency.
The Resources page originally had simple filters. But as partners added their own content across four content types, we realized users would face 40+ unique items to sift through. So we shifted to a catalog view with more granular controls, plus a toggle between card and list views to surface more options per screen.
06 · Tech Preview & next steps
A well-received first step, with clear next moves
The Tech Preview landed well as Red Hat's first foray into AI/ML platforms, and immediately made clear where we needed to rebuild the functionality we'd cut from the MVP. That expansion became the next phase of the product's growth.