Behind the scenes, with made-up data. This is a read only copy of the admin for a fictional candidate. Reports, traces, spend and chat were produced by a scripted stand-in; the evaluation timeline shows real results of real models on fictional cases. Nothing here can be changed. Back to the real site
Reverse Apply

Runs

Principal PM, AI Quality and Evaluation, Contoso AI

Run 3, started 8 Nov, 14:01 UTC, started by you. Your decision: skip

Posting against the CV

4 reports, 1 check against your terms

Posting

Principal PM, AI Quality and Evaluation

Contoso AI

Checked against

Ada Example

Received

8 Nov 2026

Reference

No. 3

claude-haiku-5.5 openrouter

Failed 810 ms

nemotron-3-super-120b-a12b openrouter

Failed 803 ms

amazon.nova-lite-v1:0 bedrock

Failed 1.1 s

baseline

Done 10% 35 ms
Result

10% match

Match: 10 of 100
Minimum requirements

4 of 4 minimum requirements not shown

“Not shown” means the CV does not show it. It does not mean the person lacks it.

How to read this

Each verdict is a model's reading of the posting against the CV, not a fact. Minimum requirements count twice as much as desirable ones, and “partly” counts half.

Minimum requirements

What the posting says it needs

  1. 1

    A record of building AI products and delivering measurable customer outcomes.

    Show it in the posting
    Not shown weight 2
    Change the verdict or the CV line

    1 of 8 requirement words appear in this line.

  2. 2

    Hands-on experience building agents and evaluating them with applied scientists.

    Show it in the posting
    Not shown weight 2
    Change the verdict or the CV line

    1 of 7 requirement words appear in this line.

  3. 3

    Experience with train and test separation, overfitting and honest reporting of results.

    Show it in the posting
    Not shown weight 2
    Change the verdict or the CV line

    1 of 8 requirement words appear in this line.

  4. 4

    Experience setting shared quality standards across several teams.

    Show it in the posting
    Not shown weight 2
    Change the verdict or the CV line

    1 of 7 requirement words appear in this line.

Nice to have

Preferred, a plus, or an example

  1. 5

    Experience with agent observability or evaluation platforms as a product owner.

    Show it in the posting
    Partly weight 1
    Change the verdict or the CV line
    Product manager for search and developer platformsCV, Ada Example

    2 of 7 requirement words appear in this line.

  2. 6

    Coaching product partners without formal authority.

    Show it in the posting
    Partly weight 1
    Change the verdict or the CV line
    Ran quarterly planning for three squads and coached two junior product managers.CV, Northfield Systems, Senior Product Manager, Data Platform

    2 of 6 requirement words appear in this line.

Open the trace

Mark a report reviewed once its verdicts are right. Reviewed reports become your labels in the evals.

The posting, marked

The posting's own text, with each requirement marked where it came from

The number after a marked sentence leads to its row in the report.

  • Met
  • Partly
  • Not shown

Minimum Requirements

- A record of building AI products and delivering measurable customer outcomes.1, Not shown

- Hands-on experience building agents and evaluating them with applied scientists.2, Not shown

- Experience with train and test separation, overfitting and honest reporting of results.3, Not shown

- Experience setting shared quality standards across several teams.4, Not shown

Preferred Qualifications

- Experience with agent observability or evaluation platforms as a product owner.5, Partial

- Coaching product partners without formal authority.6, Partial

What we offer

On site in Seattle. Salary USD 210,000 to 330,000.

The posting against your terms

Edit your terms

claude-haiku-5.5 openrouter Done Open the trace
Result

62% fit to your terms

Fit to your terms: 62 of 100
Dealbreakers

No dealbreaker conflicts

Nothing in the posting breaks a term you marked as a dealbreaker. Where the posting is silent, that is shown.

Your termVerdictWhat the posting says
Location dealbreakerThe role can be done remotely or from a city the person is willing to live in Silent
Change the verdict
The posting says nothing about it.
Seniority importantThe role matches the person's level: it is not junior or entry level Silent
Change the verdict
The posting says nothing about it.
Pay importantThe posting states a pay range, and it is at or above the person's floor Fits
Change the verdict
Salary USD 210,000 to 330,000.From the postingThe posting states it.
Domain niceThe product area is one the person wants to work in Silent
Change the verdict
The posting says nothing about it.

How many LinkedIn profiles match

A reference, not a verdict

Each number counts profiles whose text contains the search terms. It does not count people who can do the work. LinkedIn rounds and caps its counts, and the result depends on who runs the search.

No search plan yet. Draft one below. It gives you a LinkedIn search for the titles, then one that adds each requirement in turn. You run the searches and type in the counts.

Cover letter

A draft by software

Software wrote this draft from Ada Example's CV and this report. It is not written or sent by Ada Example. Each paragraph about Ada Example has to cite a CV line, or the draft is not shown. The downloaded text starts with a line that says the same.

Voice and rules the writer follows
Voice and structure:
- Open with the single most relevant proof for this role: one cited CV line, stated at the scope the line states, in the posting's own terms. Do not open with a question, with flattery, or with "I am excited".
- Then one or two more pieces of proof, each from a cited CV line. Do not turn a team's work into personal work unless the line says it was the person's own.
- Never mention what is new to the writer or what they lack. Do not address a requirement that is marked not_on_record.
- One question about the role is welcome after the proof, where it helps; never as the opening.
- End with a plain offer to talk. No sign-off flourish.
- Plain and direct, the way a person would say it to a colleague. Under 300 words, no buzzwords, at most one dash in the whole letter.

Change them under Settings.

Writes three drafts (opening with proof, with a question, with the role's problem) that only you see, then lets you compare them blind. Results so far.

Ask about this report

You are chatting with an AI model, not with Ada. Answers come from the posting, the CV and this report, and quote the CV line they rest on. Questions about anything else are declined.

Run more models

Model arena

0 votes so far

Run the same posting on two or more models. Then compare two reports with the model names hidden and vote for the better one.

Choose models to run
Models

Compare two reports blind

Your call

Private, never shared

Share

Not shared

The link shows the scores and verdicts. It never shows your notes, your terms or the traces.

The link also allows
Posting text and extra facts the models read

Posting text

Minimum Requirements
- A record of building AI products and delivering measurable customer outcomes.
- Hands-on experience building agents and evaluating them with applied scientists.
- Experience with train and test separation, overfitting and honest reporting of results.
- Experience setting shared quality standards across several teams.

Preferred Qualifications
- Experience with agent observability or evaluation platforms as a product owner.
- Coaching product partners without formal authority.

What we offer
On site in Seattle. Salary USD 210,000 to 330,000.

Delete this run with its results, chat, letters and traces. This cannot be undone.