CVFriend · job applications made easy
I am an engineer and a marketer. I tried every CV builder and chatbot I could find and they all produced the same thing: a fluent, confident paragraph that says nothing, the kind a recruiter learns to skim past in two seconds. So I spent months building one that writes like me instead, and refuses to ship anything that does not.
No signup needed for the score. See the difference first.
Free to start · no card required · platform-funded · GDPR-compliant, hosted in the EU
Senior Performance Marketer
Klarna · Stockholm
The tell
These are the exact phrases I kept reading, in builder after builder. Read the paragraph once and you cannot unsee it. A recruiter reads it dozens of times a day.
This is what a language model writes when you ask it to sound impressive. It reads fluent and dies on the recruiter's screen. I treat every one of these as a hard fail.
Why I built it
I am an engineer and a marketer. When I was let go, I joined the part of the job market nobody enjoys: writing a fresh CV for every role, tuned to whatever the screening software wanted to see, then sending it into the void. Half the time the answer was no. Plenty of the time there was no answer at all. It was soul-consuming.
I tried the CV services and I tried talking straight to the chatbots. Nothing wrote anything I would actually put my name on. So I started building my own, small at first, just enough to take the worst of the busywork off my plate. Over months of applying every day and fixing whatever annoyed me, it grew into something that did the whole job: it found the roles, read the fit, and wrote each application in a voice that was mine.
In one interview the recruiter complimented my application, then in the same breath complained about every machine-written cover letter they were drowning in. I sat there knowing my system had written every word of mine.
From the first line of code to a signed full-time contract took three months. The night I powered the machine down, friends started asking me to help with their CVs. So I turned the whole process into CVFriend, the version you are looking at now.
What I found
I did not start with a product. I started with a list of everything wrong with a machine-written application. Here is the list, and what I built for each item on it.
What it gets wrong
It reads like a machine.
A voice check runs on every export. Carry the mechanical tells and the file does not leave the app until they are gone.
voice checkWhat it gets wrong
It invents your career.
A provenance guard ties every number and claim back to something you actually told me. The system cannot fabricate a metric to make a line read better.
provenanceWhat it gets wrong
Screening software bins it.
A deterministic score from 0 to 100, with the exact keywords you are missing, before you send. You watch the number move as you close each gap.
ATS scoreWhat it gets wrong
It sounds generic, not like you.
It learns your voice from your own writing and gets closer to it with each application, instead of flattening everyone into the same register.
your voiceWhat it gets wrong
The same CV goes everywhere.
Each application is rewritten for the specific role, with a fit and gap read on the job before you write a word.
per roleWhat it gets wrong
You still fix the robot voice yourself.
Other tools hand you a draft and tell you to clean it up. I block the slop at the source, so there is nothing to clean up.
at the sourceThat is the hybrid. The speed of a generated draft, held to the bar of a hand-written one, with checks no generic builder runs.
Left is what a generic builder produces. Right is what I ship. Read both out loud. That is the test, and you can run it yourself.
"As a highly passionate and results-driven professional, I am excited to leverage my extensive expertise — to drive impactful results and synergize cross-functional initiatives. I am a dynamic self-starter who thrives in fast-paced environments — delivering value, fostering growth, and pushing boundaries — and my passion for excellence is unparalleled."
em-dash spray · passionate / results-driven · rule-of-three · zero numbers
Performance marketer with eight years owning paid acquisition across Google, Meta and TikTok. I ran a 40 MSEK annual budget at Nordic365 and cut blended CAC by a third in 18 months. I built a marketing-mix model with finance that moved 6 MSEK to higher-return channels without losing volume.
specific · quantified · human · keyword-aligned to the job
The right panel is a real CVFriend output. The left is a generic-builder sample, shown for comparison. This is the bar every export has to clear.
Free to start · no card required · cancel anytime
Under the hood
Most tools send your details to a model once and hand you whatever comes back. Here, every application runs through four or more separate verification and rewriting passes, each one a proven skill with strict rules, built to push the quality up and keep invented claims out. Three of those passes are checks no generic builder runs.
A deterministic voice check runs on every export. If the writing carries the mechanical tells of machine prose, the file does not leave the app until they are gone. Enforced in software, not left to you.
Every CV gets a score from 0 to 100, with the exact keywords you are missing, before you send. You watch the number move as you close each gap.
Every number and claim has to trace back to something you told me. The system cannot fabricate a metric to make a sentence read better. On a CV, that is the line between confident and dangerous.
The compounding part
Your profile is not a form you fill in once. It is a living record of your career that the system keeps learning from. Every data point you add, every question you answer, every application it writes teaches it more about you, so the next generation is faster and asks less of you.
It studies the edits you make and the cover letters you upload in your own words, then moves its writing toward your style instead of a generic one. The more you use it, the more each application reads like you wrote it on your best day.
What it learns from
The result: each application is quicker to produce and reads more like you than the last.
Live today
Everything below runs right now, and I use it for my own applications. The three checks marked in blue are the parts a generic builder does not have.
The gate every export has to clear before it leaves the app.
Deterministic 0 to 100, with the exact keywords to add.
Numbers trace to your profile. Nothing gets invented.
Upload your current CV once. I pre-fill it; you correct what is off.
Live roles from Platsbanken for Swedish users, matched to your profile. Outside Sweden, paste any job link and the rest works the same.
A straight read on whether the role is worth your time, plus what you match and what you are missing, before you write a word.
Rewritten per role in your voice, with the score attached.
A real personligt brev or English cover letter, not a template fill.
An ATS-safe .docx for the portal and a styled .pdf for the recruiter.
I am still building. The most recent addition is a one-click monthly report of the jobs you applied to, formatted for the Arbetsförmedlingen activity report.
The problem
Send the same CV everywhere and let volume carry it. Fast, and the reason most applications never get read.
Rewrite for every role and burn an evening per application. It works. Almost nobody keeps it up past the first week.
Tailored to each role like the slow way, fast like the lazy way, and held to a bar neither clears: it reads like you wrote it, and it is built to pass the software that screens you out.
Upload your current CV. I extract and pre-fill it. Correct what is off and you are done.
In Sweden, a feed of live Platsbanken roles matched to your profile, refreshed every morning. Anywhere else, paste a job link and go straight to tailoring. No trawling boards.
One click writes your CV for that role in your voice, gaps flagged and every line voice-checked. Add a cover letter if you want one.
An ATS-safe .docx for the portal and a styled .pdf for the recruiter, ready the moment your CV passes the voice check.
The bigger picture
Companies have handed the first decision to software. Screening systems and machine filters now decide who a human ever sees. The people who lose are not the unqualified ones. They are the ones who would not stoop to slop and did not have an evening to hand-tailor every application.
My job is to even that out. Make it ordinary for a qualified person to send an application that sounds like them and clears the machine on the other side. I start with the Nordic market I know best, with Swedish personligt brev conventions handled properly and the local job sources crawled every morning. The standard travels from there.
Your balance is measured in applications, so you always know where you stand. Start free, upgrade when the feed starts paying off. No surprise bills, cancel anytime.
Free · Starter 119 kr/mo · Pro 239 kr/mo
GDPR · EU-hosted · your CV never trains a model · cancel anytime
Build your profile, score one CV, run the voice check. Free, and it takes about five minutes.
Free to start · no card required · cancel anytime