Wisecube: AI evaluation and biomedical research tools.

Our work with Wisecube covered two different products: Pythia, for checking AI responses against reference text, and Orpheus, for exploring connected biomedical information.
Wisecube logo
Engagement
Client work
Our part
Research product interfaces
Our contribution

What we worked on

A clear account of the work we did, kept separate from the wider product.

  1. 01

    Pythia: evaluate the user’s own text

    Connected the demo to evaluate the reference, response and question the user enters. Also handled submissions where the question is left blank.

  2. 02

    Pythia: inspect the returned claims

    Connected returned evaluations to the claim-by-claim results view, showing whether the reference supports, contradicts or does not establish each claim.

  3. 03

    Orpheus: explore graph details

    Added detail views for selected graph nodes and connections, plus relevant insights from the starting node.

Pythia reference text, AI response and corresponding claim checks shown togetherView full screenshot
Original Pythia demo with fictional inputs and results. The response invents unlimited storage; the corresponding claim is marked as a contradiction. This illustrates the interface, not model accuracy.