Pienso Match
Powerful text search that matches meaning, not just key words
- UI
- UX
- AI/ML
- 0->1
Watch the full-length video of vector search in use here.
Match is a semantic search tool that gives users close control for finding sentences and documents with similar meaning.
Team
Design Will Crum
Development Brian Cort, Felipe Balduino Cassar
Background
We were testing out a new architecture for training-free, “zero-shot” text classification — and it needed a UI.
Objective
Develop the MVP for Explore, a document discovery tool built on a vector embedding-based architecture of semantic similarity.
Process
This was a 0-1 flagship feature MVP stand-up. Briefed in May 2022, MVP in dev by EOY, with continued refinements into 2024, juggled with other priorities.
Challenges
- Open-ended brief – We were testing tech, not solving a user problem
- Hidden complexity – Everyone knows text search, but no one knows vector distance. Should they?
- DistractGPT – When ChatGPT came out, leadership's vision for Explore's UX was shaken.
Background
- Advanced but accessible – It let users compose complex queries, but kept results digestible.
- A testable MVP – The search UX was crucial to understanding and debugging LLM vector embedding.
- User adoption – Existing customers used it to quickly query last week's customer calls for new issues