← Back to work
LiveAgent

Job-Hunt Agent

Scores live job ads against my CV, then drafts the application.

What this shows: I can break a tedious manual workflow into steps an agent runs end to end, with hard filters and me making the final call.

Outcome

A morning of scrolling job boards becomes a handful of roles actually worth applying to, each with a resume and cover letter already drafted for it.

Proof

One real run: 27 roles evaluated in depth, the top ten all scored 4.6 or higher against my CV, and four hit 4.8. Every verdict came with a written report.

Takeaway

Hard filters beat clever scoring. Skip is a verdict too.

The pipeline, end to end

From a job posting to a ready-to-send applicationLIVE
one real run1Scan30 sources,filtered to fit2Extractread the fulljob posting3Verifyreal role,still live?4Scorefit 1 to 5against my CV5Tailorgaps + anATS-ready CV6Preparedraft apply,track to send27 roles evaluatedtop 10 scored 4.6+4 fits at 4.8

At a glance

30
Sources watched
27
Roles evaluated in one run
4
Top fits, scored 4.8 out of 5
191
Listings tracked so far

How it works

  1. 1

    Scan the boards

    Pulls fresh listings from 30 tracked sources, company boards plus the big job sites, and filters to the roles I’m actually after.

  2. 2

    Read the posting

    Extracts the full job description from each link, even the awkward single-page application portals.

  3. 3

    Verify it’s real

    Checks the role is genuine and still live, not a stale repost or a ghost job, before I spend time on it.

  4. 4

    Score the fit

    Rates each role one to five against my real CV, and flags where I’m strong and where I’m short.

  5. 5

    Tailor the CV

    Builds an ATS-friendly CV for that exact role, pulling in the keywords the posting actually asks for.

  6. 6

    Prepare to send

    Drafts the application answers, then logs and tracks the role so nothing slips through.

Scored against my CV

Every listing gets a score out of five and a verdict. Only the top ones reach me:

AMP

AI Product Owner

4.8apply

Heidi Health

Product Manager, Intelligence

4.8apply

EY

Decision Modelling, AI & Decision Analytics

4.8apply

Glean

AI Outcomes Manager, ANZ

4.8apply
real output · one run

Problem

Job hunting is a volume game that rewards tailoring, and doing both by hand doesn't scale. I wanted the boring filtering and the first drafts handled, with me kept in the loop.

Approach

It sweeps 30 tracked sources, company boards plus the big job sites, and filters to the roles I actually want. Each surviving posting is read in full and checked that it's genuine and still live, then scored one to five against my real CV. Strong matches get an ATS-friendly CV built for that exact role, the application answers are drafted and tracked, and I review and send.

Eval results

From a real run on 11 June 2026: 27 roles evaluated, the top ten all scored 4.6 or higher, and four hit 4.8: an AI product owner, a product manager in health AI, a decision-analytics consulting role, and an AI outcomes manager. Each verdict arrives as a written report I review before anything is sent. Next up: a proper eval of the scoring itself, the agent's ranking against my judgement.

What broke

The first scorer over-weighted keywords and loved any ad that said the word AI. Adding seniority and context checks pulled it back toward sane rankings.

Curious how it scores a role? Ask.

🥬Kale Bot

Ask me anything about Cael and his projects. I answer from real sources, and I will tell you if I do not know.