Converting a complex CLI to an intuitive GUI
Designing a visual web app to make dataset anomaly detection accessible
Context
DaVe is a web app for detecting anomalies in datasets. I co-designed the full experience in a cross-functional team, from shaping the MVP to refining the final interface.
Team: PM, 2 designers, 1 technical writer, 1 QA, ~4 software engineers, ~3 data engineers, and strong CEO involvement.
Business case: SigTech mainly bought raw data, cleaning and aggregating it for data-consumers. If we gave data owners a tool to rapidly validate and improve their datasets before selling, it wouldn't just create a new revenue stream — it would also save our data team time and improve the experience for data-consumers, creating a virtuous cycle for the whole business.
('DaVe' = Data Validation Engine!)
Requirements gathering
Collaborating with the PM and CEO, I captured the core requirements in a written spec — a crucial step in surfacing hidden differences in our assumptions about our users, especially their level of technical knowledge.
Next, I wireframed the end-to-end journey that guided engineers' plans for the build. Early concepts included letting users edit anomalies as well as detect them, but we quickly realised this was too ambitious for an MVP to be launched in only weeks. We scaled back to detection-only.
Testing core usability
We created a clickable prototype of the core flow, enabling realistic, hands-on reviews and faster alignment across the team.
New brand. New design system.
This was the product's first product UI expression of our recent rebrand, so visual clarity and brand alignment were key. To support this, we designed, from scratch, a library of Figma components using atomic design principles.
Edge cases and details
With the core flow and aesthetic validated, we moved on to designing all edge cases, minor UIs, and UI states. We stayed at least one sprint ahead by working closely with engineers throughout.
Project reflections
Wins
- Strong cross-functional collaboration across design, engineering, product, and QA.
- Fast development of first rebranded product UI.
- Early investment in a design system paid off, accelerating the next project.
- Saw initial traction (but ultimately neglected as company pivoted to AI with the release of ChatGPT plugins).
Learnings
- Some wireframe reviews lacked thoroughness — e.g. the sequence of backend calls involved was missed — leading to avoidable rework during UI design.
- The initial MVP scope was too ambitious, causing inefficiencies as wireframes had to be revised midstream.
