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

Does this job fit Ada?

Give the link to a job posting. Each requirement is checked against Ada Example's CV, and the CV line behind every verdict is shown. Then you can ask questions about the match, and a draft cover letter comes with it.

Check a job posting

Give the link

Checks are paused for now. Please come back later.

You can still read the example report.

What a report looks like

Made-up example: fictional candidate, fictional posting
Posting

Senior product manager, search

Northwind Labs (fictional)

Checked against

Ada Example (fictional)

Result

61% match

Match: 61 of 100
Minimum requirements

1 of 4 minimum requirements not shown

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

Minimum requirements

What the posting says it needs

  1. 1

    Owned a search or retrieval product used by developers

    Met
    Product owner for the site search API used by 2,000 developers.CV, Example Corp
  2. 2

    Defined relevance metrics and ran offline evaluations

    Partly
    Set up weekly relevance reviews with the data team.CV, Example Corp

    Reviews are named, but no metrics or offline evaluation are on record.

  3. 3

    Experience with usage-based pricing

    Not shown

    Nothing in the CV mentions pricing.

  4. 4

    Led a cross-functional team of engineers and designers

    Met
    Led a team of eight across engineering, design and data.CV, Example Corp

Nice to have

Preferred, a plus, or an example

  1. 5

    Familiarity with embeddings and reranking

    Partly
    Shipped semantic search using an off-the-shelf embedding model.CV, Side project

    Embeddings are named; reranking is not.

How it works
  1. GiveGive the link to a job posting. If the page cannot be read, you can upload it as a PDF. No account and no e-mail address.
  2. ReadA model lists what the posting asks for and sorts it into minimum and desirable requirements.
  3. CheckEach requirement is checked against the CV and stamped: met, partly, or not shown. The CV line behind the verdict is shown next to it.
  4. AskAsk follow-up questions. The chat answers from the posting, the CV and the report, and can argue against the match if you ask it to.
  5. DraftGet a draft cover letter in which each paragraph about the candidate cites a CV line.

What it will not do
  • Decide for you. A verdict is a model's reading of two texts, and it can be wrong. The CV line is there so you can check it.
  • Say “no”. “Not shown” means the CV does not show it, not that Ada Example cannot do it.
  • Speak for Ada Example. Software writes the report and the letter. Ada Example does not write or send them.
  • Answer questions outside the posting, the CV and the report.
  • Keep your text for long: 30 days, or until you delete the report.

How it is built

Each step (read the posting, judge the evidence, answer, write) is a small agent built with Pydantic AI.

Every model call is recorded as an OpenTelemetry trace: the prompt, the answer, the time and the cost.

Evals check the agents against test cases with known answers, against reports the owner corrected by hand, and with a second model that judges the cited CV lines.

Open the dashboards