ReactiveCore · AI / Knowledge Engineering · Ontology Design

AI Metadata Factory

Designed a simplified web flow, markup sequence, and visualization system that allowed non-expert administrators — not just Knowledge Engineers — to create and visualize OWL ontologies from client files.

ReactiveCore AI Metadata Factory — overview screenshot
Client
ReactiveCore
Domain
AI / Knowledge Engineering
Employer
Cognizant
Deliverable
Tested Prototype
My Role
UX Designer / Researcher

The Problem

ReactiveCore's AI platform relied on Knowledge Engineers (KEs) to take client data files and manually mark them up to produce OWL ontology files — a highly specialized, time-intensive process. This created a bottleneck: only experts could perform this work, slowing down the pipeline and limiting scale.

The challenge was to design a tool that would allow less-specialized administrators to perform the markup and visualization workflow — without requiring deep knowledge engineering expertise — while still producing accurate, usable OWL output for further diagnostic use.

My Role & Approach

I researched and designed the AI Metadata Factory application, working closely with ReactiveCore's chief ontologist to understand the markup and ontology creation process in depth before translating it into a simplified, guided web interface.

The core design challenge: Ontology creation is inherently complex. The goal wasn't to hide that complexity — it was to sequence it into steps an admin could follow confidently, with the right scaffolding at each stage.

Expert mental model research and task analysis
Expert mental model mapping — translating Knowledge Engineer workflow into admin-friendly steps
  • Deep research sessions with the chief ontologist to map the expert mental model
  • Task analysis of the full KE workflow: file intake, markup, OWL generation, visualization
  • Designed a simplified step-by-step web flow for non-KE administrators
  • Created a guided markup sequence with contextual guidance and validation feedback
  • Designed an ontology visualization interface for reviewing and diagnosing the output
  • Conducted prototype testing with users at multiple fidelity stages

Design Highlights

Simplified markup sequence: Rather than exposing the full complexity of OWL markup upfront, the interface walked administrators through a staged process — each step building on the last, with clear progress indication and the ability to review and revise.

Simplified markup sequence — step-by-step guided flow
Simplified markup sequence — staged wizard with progress indication and contextual guidance

Ontology visualization: Once markup was complete, the resulting ontology was rendered as an interactive visual graph — allowing admins and KEs to review relationships, spot gaps, and navigate the structure without reading raw OWL files.

OWL ontology visualization — interactive graph
Ontology visualization — interactive graph enabling non-expert review of OWL structure

Accessibility to non-experts: Every UI decision was evaluated against the question: "Could an intelligent but non-specialist admin complete this step without calling a Knowledge Engineer?" By the end of testing, the answer was consistently yes.

Outcomes

Prototype tested with end users and received strongly positive feedback — administrators could complete markup workflows that previously required a Knowledge Engineer.
Replaced an unusable Jupyter Notebook interface with a guided step-by-step web flow, validated directly with the head ontologist until the sequence matched how the process actually worked.
Reduced dependency on specialist Knowledge Engineers for routine markup tasks — the core bottleneck the tool was designed to solve.
Ontology visualization made accessible to the broader team — non-experts could review OWL structure without reading raw files. Prototype was user-tested and validated; not moved to production during this engagement.

Skills & Tools

AI / ML UX Ontology Visualization Knowledge Engineering Task Flow Design Expert Research Prototyping User Testing Figma
← Warner Media HBO Next: PepsiCo →