Redesigning Decision-Making for a Global Supply Chain

A global supply chain team needed faster, more confident decisions.

I led design and product strategy for Nvidia's supply-chain intelligence platform, an enterprise AI analytics tool managing $2B+ in annual transactions. It served global teams making high-stakes operational decisions on AI-generated insights and real-time data.

The existing interface was flat and unstructured; information was presented without hierarchy, forcing users to mentally organize data before they could act on it. This slowed decision-making and increased error rates, particularly for cross-functional teams who needed different views of the same data.

Enterprise structure meets AI-powered intelligence.

SAP Fiori as the structural foundation. I selected SAP Fiori as the design framework for two reasons: it is the enterprise standard that supply chain teams already understand, and its structured dashboard patterns (Overview Pages, Object Pages, List Reports) map directly to supply chain decision-making workflows.

Information architecture overhaul. I restructured the platform's IA from a flat list of features into a task-oriented hierarchy. Users no longer navigated by feature name; they navigated by decision type: "What needs my attention?", "What is the current status?", "What action should I take?"

AI integration patterns. I designed interaction patterns for AI-generated recommendations that maintained user agency. The AI surfaced insights, but humans made the final decisions, with clear confidence indicators and the ability to drill into the underlying data.

From ten filters to one prompt. Building a query used to mean configuring up to ten filters by hand. I replaced that with a conversational flow: users describe what they need, the AI proposes the query and runs the projection against live platform data. The gain was not speed — it was confidence. Users asked follow-up questions instead of accepting the first answer, tracing a component down to its class and origin without leaving the conversation.

The restructure was validated, not assumed. In internal usability testing with 10 participants — marketing users, not engineers — 9 found the new flow easier and less error-prone than the filter-based query it replaced. The tenth did not: they had built a mechanical habit around the old filters and never realized the same query could now be expressed as a prompt. That was the most useful finding of the round — a discoverability gap in how the interface advertised its own capability, not a failure of the flow itself.

The deep-reasoning, guardrails, conversation-history, and data-flywheel screens that bring these patterns to life are withheld under NDA. The interaction design is described above; the full work is available to walk through on request under a mutual NDA.

Beyond design: functional product ownership.

On this project, I operate beyond the traditional designer's scope. Acting as Product Owner on the design side, I write user stories with acceptance criteria so engineering teams understand the intent behind each design decision. I run design QA on the build, facilitate weekly syncs with 5+ engineers including AI/ML specialists, and govern the design system to ensure Atomic Design compliance across the product.

I also mentor 2 junior designers who contribute to the component library and product screens, enabling them to work autonomously within the design system I established.