Question 1

What problem does the Clay JEV people ranker solve?

Accepted Answer

Clay's deterministic filters match broad titles such as founder, but also catch founding investors, founding employees and Founder's Office roles. The workflow uses Jev to make the semantic call on whether each person actually holds an operating Founder or Co-Founder role.

Question 2

How do the Clay CLI and Jev split the work?

Accepted Answer

The official Clay CLI retrieves people who match deterministic filters like job title, company size, industry, seniority and location. Jev then judges the remaining semantic ambiguity with a Choice classification and a Noul probability, and Python applies the thresholds.

Question 3

Which TypeSafe primitives does the workflow use?

Accepted Answer

Two: Choice classifies the candidate as an operating founder, investor or board member, founding employee, founder-support role or unclear; Noul estimates whether the candidate clearly and currently holds an operating Founder or Co-Founder role.

Question 4

What did the measured founder-ranking test show?

Accepted Answer

On 2026-09-21 the workflow evaluated 577 Clay candidates from a 500-person US software-founder target: 472 were Jev-qualified and 105 rejected before enrichment, an 81.8% qualification yield, with 20.115 seconds of scoring from cached Clay data.

Question 5

What do the 0.90 and 0.80 thresholds mean?

Accepted Answer

They are operating policies, not accuracy guarantees: 0.90 for the current-founder Noul and 0.80 for the operating-founder question. The README explicitly says the 0.90 threshold is not proof of 90% precision and that production use needs a human-labeled evaluation set.

Question 6

What did the test run cost and where did it stop?

Accepted Answer

The run consumed 490,395 Jev input tokens and 53,648 output tokens, an estimated $0.0206 of input charge. It stopped at 472 of the 500-person goal after a Clay continuation request returned HTTP 504, though the Jev scoring stage completed for all 577 candidates.

Question 7

Who is it for and what does it not do?

Accepted Answer

It targets GTM engineers, RevOps teams, sales operations, CROs, BDMs, SDRs, recruiters and developers building programmable lead qualification. It does not replace Clay: Clay retrieves the market, Jev judges ambiguous fit, and Python controls the policy before enrichment.

Clay JEV People Ranker

Highlights

Questions about this entry

From the Jev guides

From the Jev AI model community