Skip to content
Jacob Clifton
← All work

Katch · 2023 to present

Katch Verified

Real-time brand-safety and brand-fit scoring across a creator's entire body of work, so a partnership decision stops being a manual review.

Role · Co-founder and CTO. System design and modeling.

Visit

The problem

Brands and agencies commit real budget to creators on the basis of a follower count and a spot check of the last few posts. The risk lives in the archive, and the fit question is not binary: a creator can be perfectly safe and still wrong for the brand. Both answers have to arrive before the deal is signed, which means the review has to be automatic.

Approach

  1. 01

    Extended the semantic analysis layer from long-form film and television to short-form social, music, and books, so one representation covers every format a creator publishes in.

  2. 02

    Scored fit and risk as separate axes, because collapsing them into a single number destroys exactly the information a brand manager needs.

  3. 03

    Built the scoring path to run at request time against a full back catalog rather than a sampled window.

Outcome

Live product, used by brand and agency teams for creator vetting and partnership diligence.

Stack

PythonLLMsContent classificationRankingReact