Privacy

Privacy policy

Radpul stores the raw material of your career: your CV, what you did at each job, and the resumes built from it. This page says plainly what is collected, who else it reaches, and what you can ask for. It describes the system as it is built today, not as it is planned.

Operator

Legal name
Arda YILDIZ
Address
Çankaya / Ankara, Türkiye
Contact
[email protected]
Courts
Ankara
In force since
20 November 2026

1. Who is responsible

The operator named above decides why and how your personal data is processed, and is who you contact to exercise any of the rights in section 8.

2. What is collected

Account: your email address, a hash of your password (never the password itself), whether the address is verified, the invite code you joined with, if you joined with one, and the language you use Radpul in (English or Turkish), so its email reaches you in that language and it opens in that language when you sign in on a browser where you have not chosen one. Until you choose one with the language switcher, it is the language Radpul showed you when you signed up or, for an older account, the next time you opened Radpul signed in: your browser's language if Radpul offers it, otherwise English. Career content: the CV file you upload and the text extracted from it; the work experience, projects, education, skills, languages, certifications and evidence you approve into your Career Memory; what you write in Write with AI and the suggestions made from it, including the ones you reject; job postings you paste in; and the resumes generated from them, including the PDF and LaTeX files. Photo: if you add a profile photo, the file you upload is turned into two smaller JPEG copies, a portrait and a square, and then deleted. The copies are new images, so the location, camera and time details a phone records inside a photo are not carried over. Your resumes print one of them unless you turn the photo off for a resume. Removing the photo deletes both. References: if you add a referee, you are supplying another person's contact details. You are responsible for having their permission before entering them. Sessions: for each sign-in, the browser's user-agent string and IP address, so that you can see and revoke your own active sessions. Sign-up source: if the visit that led to your sign-up came through a link carrying campaign tags (utm_source and the like) or from another website, those tags, that website's name (not the full address of the page) and the first page you opened on Radpul. They are kept with your account so Radpul can count which channels bring people, are part of your data export, and are not shared with anyone. Product analytics: the pages you open on Radpul, the address of the site that linked you there (without the rest of its address) and the steps you complete, which PostHog counts without cookies; and, only if you accept them, a PostHog identifier for your browser, where you click and scroll, how fast pages load, the place your visit came from, recordings of your sessions with all text hidden, and your account ID once you are signed in. Section 10 describes each. Crash reports: when the application fails, a report is sent to the error-monitoring provider. It is configured to exclude request and response bodies and component properties, which is where CV text, prompts and passwords would otherwise appear.

3. Why it is processed

Each purpose below rests on one of the grounds listed in Article 5(2) of the Turkish Personal Data Protection Law (KVKK, Law No. 6698) or, for special categories of data, Article 6(3). To provide the service you asked for (creating and running your account, signing you in, parsing a CV, turning what you write in Write with AI into suggested Career Memory entries, analysing a job posting, generating a resume, sending verification and security email): the processing is necessary to perform the contract formed by the Terms of service (Article 5(2)(c)). To keep accounts and the service safe (rate limiting, session management, detecting reuse of a stolen refresh token, the security record of sign-ins and account changes): Radpul's legal duty to keep personal data secure (Article 5(2)(ç) with Article 12) and its legitimate interest in a service nobody can take over (Article 5(2)(f)). To fix faults (crash reports, with personal content excluded) and to recover from a failure (a nightly encrypted backup of the database, kept for 14 days, see section 7): Radpul's legitimate interest in a service that works and does not lose your data (Article 5(2)(f)). To count which channels bring sign-ups and how the site is used without recognising you (the sign-up source, Cloudflare Web Analytics and PostHog's counts without cookies, see section 10): Radpul's legitimate interest in knowing where its users come from and which steps work, using only the data described there (Article 5(2)(f)). To recognise your browser across visits, analyse your use of Radpul in more detail and record sessions with PostHog: your explicit consent, given or refused in the analytics question and withdrawable at any time (Article 5(1)). To see how far accounts get and what their AI use costs, from records Radpul already keeps for the purposes above (when an account signed up, verified its email address, gave the consent a CV import needs, uploaded a CV, had it ready for review and applied it to Career Memory, analysed a job posting, made its first resume and PDF, and came back on a later day; and the AI calls it made, with their tokens and their cost at the provider's list price): Radpul's legitimate interest in knowing which steps work and what the service costs to run (Article 5(2)(f)). The operator sees this for groups of accounts, by sign-up source and by sign-up week, and for a single account on its page in Radpul's admin tools (section 9), to help that account when it asks for help or gets stuck and to notice unusual AI cost. Nothing extra is collected for it: the one thing kept for it is the approximate date an account first came back, kept until the account is deleted, so that the count does not change when old session records are deleted (section 7). Nothing is sent outside Radpul, no CV or Career Memory content is read, and it is not used to decide anything about you. To process any health, religion, union membership or other special category of data a CV you upload contains: your explicit consent, which Radpul asks for separately before a CV import and which you can withdraw at any time (Article 6(3)(a)). See section 6. To answer a request under section 8 and to establish or defend a legal claim: Article 5(2)(ç) and (e). Radpul collects this data electronically: from what you enter and upload, partly by automated means (the CV parser and the AI provider), and from your browser while you use the site. Nothing is processed for advertising or profiling, and your data is not sold.

4. Who else processes it

Radpul runs on infrastructure operated by others. Each of them processes data only to deliver its part of the service: Railway: application servers, background workers, the PostgreSQL database and Redis. Hosted in the United States (US West). Cloudflare: the website itself, R2 object storage holding uploaded CV files, profile photos, generated PDFs and the encrypted nightly database backups (see section 7), and Web Analytics, which measures page loads (see section 10). Anthropic: the Claude API, which receives CV text (and a PDF CV of up to six pages as the file itself), job postings, what you write in Write with AI, and Career Memory content in order to extract and generate. See section 6. Resend: delivery of verification, password-reset, security and account email. Receives your email address and the email itself. Sending runs in Ireland; Resend's main processing takes place in the United States. Sentry: error monitoring, configured to exclude personal content. Hosted in the European Union, in Germany. Hostinger: the mailbox behind the contact address above, which receives the messages you send to it. PostHog: product analytics and, if you accept it, session replay, run in PostHog's European cloud in Germany and reached through an address on radpul.app. Receives the pages you open, the steps you complete, the site that linked you to Radpul (its address only, without the page or query), your IP address (used as section 10 describes, then discarded) and your browser, operating system, device type, screen size, language and time zone; if you accept analytics, also where you click and scroll and how fast pages load and, if you accept session replay, recordings of your sessions, with your Radpul account ID once you are signed in. Never your name, email address or what you wrote. See section 10.

5. Where your data is

The database is kept in the United States, with Railway. The files you upload and the encrypted backups are stored by Cloudflare in its Eastern Europe storage location, and Cloudflare serves the website from its network around the world. AI processing also takes place outside Türkiye: Anthropic stores what it receives in the United States, and may run the processing itself in any country where it operates. Using Radpul therefore involves transferring your personal data abroad. This is a property of the architecture, not an option that can be switched off per user.

6. AI processing

CV text, job postings, what you write in Write with AI, and the parts of your Career Memory needed for a given operation are sent to Anthropic's Claude API. A PDF CV of up to six pages is also sent as the file itself, so the model can see its layout; whatever the file holds goes with it, a photo in it included. Your profile photo never is. A CV can hold details an application rarely needs, such as your health, religion, union membership or a photo of you. The law treats most of these as special categories of personal data, so before your first CV import Radpul asks, separately from this notice, for your explicit consent to process any such details the file holds. If you refuse, you can still enter your career by hand; a CV import needs this consent. You can withdraw your consent at any time under Account; Radpul then deletes the CV files you imported and the text extracted from them. Radpul's own checks compare what comes back about you with what you provided, and discard a line that adds an employer, date, number or technology they cannot find there. The checks are automatic and can miss a case, so read each line before you approve or use it. Those checks govern what comes back. What Anthropic does with what is sent is governed by its commercial terms for the API. Under them, Anthropic does not use this content or the responses to train its models, and deletes both within 30 days. One exception applies: if Anthropic's automated safety systems flag a request as a possible violation of its usage policy, that request and its response may be kept for up to two years for review, and the safety classification scores for up to seven years. Anthropic may also keep data longer where the law requires it. Radpul does not use your content to train any model of its own.

7. How long it is kept

Your account and its contents are kept until you delete them, unless the email address is never verified, as described below. Delete account, under Account, schedules the deletion: seven days later your account and everything in it are deleted, including the files in storage. Until then you can sign in, but only to cancel. An account whose email address is never verified is also deleted, unless you sign in to it with Google: if it is still unverified 30 days after sign-up, one email reminds you, and if it is still unverified seven days after that reminder, the account and everything in it are deleted. Either way, a last email to your address says that the account was deleted and why. Apart from the backups described below, what remains afterwards cannot be traced back to you: a record of each AI operation (its type, the model and the tokens it used, never its content); a record that an account was deleted, without saying whose, kept for three years; the security record of sign-ins and account changes, without saying whose, until it is a year old; records of background work, without saying whose, until 90 days after the work finished; and PostHog's counts made without cookies (section 10). The one exception is your account ID, which Radpul keeps until PostHog has accepted the request to delete your analytics and for 30 days after that. Radpul keeps it so that it can check that PostHog deleted your analytics and, if a backup made before your account was deleted is ever restored, delete your account again. Then it deletes the ID. While your account exists, records about it that have served their purpose are deleted on their own. The record of a sign-in session, with your browser's user-agent string and IP address, is deleted 30 days after the session ends, and the record of a link to verify your email address or reset your password 7 days after the link was used or expired. The security record of sign-ins and account changes is deleted after one year, except the record of a consent you gave or withdrew for CV imports, which is kept as long as your account exists. Records of background work, such as a CV import or a resume being written (with no content), are deleted 90 days after the work finished. Each round of these deletions leaves a record of how many items were deleted, without saying whose, which is kept for three years. Every night Radpul makes a backup copy of its database, which holds your account and everything in it apart from the files themselves (uploaded CVs, photos and generated PDFs). Each backup is encrypted before it leaves Radpul's servers and is deleted automatically 14 days after it is made. Anything you delete, a whole account included, therefore stays in the backups made before the deletion for up to 14 days, and the storage provider can take a day or so longer to remove an expired backup. The backups are used only to restore the service after a failure and to test that they can be restored.

8. Your rights

Under Article 11 of the KVKK you may ask Radpul whether your personal data is processed and, if it is, for information about it; what it is processed for and whether it is used for that; who in Türkiye or abroad it has been transferred to; for incomplete or inaccurate data to be corrected; for it to be deleted or destroyed; for a correction or deletion to be passed on to those it was transferred to; to object to a result against you that comes solely from automated analysis; and for compensation if unlawful processing has caused you damage. To make a request, email the contact address above from the email address registered to your Radpul account. Radpul answers within 30 days, free of charge. If your request is refused, you find the answer inadequate, or no answer comes in time, you may complain to the Personal Data Protection Board (Kişisel Verileri Koruma Kurulu) within 30 days of learning the answer and in any case within 60 days of your request. Download your data, under Account, gives you a copy at any time: your records as Radpul stores them, with the CVs you uploaded, your photo and your latest resume PDFs. Delete account, under Account, deletes everything as section 7 describes.

9. How it is protected

Traffic is served over HTTPS only, with HSTS. Passwords are stored hashed, never in readable form. Session tokens live in HttpOnly cookies the page's own JavaScript cannot read, and are rotated on refresh; reuse of an old token revokes the whole session. Uploaded files sit in private object storage and are reachable only through short-lived signed links, or, for your profile photo, through Radpul itself while you are signed in. The nightly database backups sit in a separate private bucket and can be decrypted only with a key that the operator alone holds, which is kept neither on Radpul's servers nor with Cloudflare. To check that a backup can be restored, the operator from time to time decrypts one on their own computer, counts the accounts in it without opening their contents, and then deletes that copy. Radpul's admin tools show account details and activity only: for example, when you first reached each step described in section 3 and what your AI calls cost. They never show CV or Career Memory content, and nobody signs in as you. The operator can reach the database directly to run and repair the service, and does not open your content unless you ask for help with it.

10. Cookies and analytics

Radpul sets four first-party cookies that are strictly necessary and need no consent: radpul_access, which carries your session for 15 minutes, and radpul_refresh, which renews it for 7 days, or 30 if you chose to stay signed in; radpul_csrf, which carries a token that prevents another site from acting as you and is removed when you close the browser; and radpul_signed_in, which holds only the value "1" for as long as radpul_refresh lasts, so the site knows to open your workspace instead of first showing its public pages. It also keeps a few things in your browser's local storage so that they survive a reload: your light or dark theme; the language you chose with the language switcher, if you chose one; your answer to the analytics question below and the date you gave it; and, kept separately for each account signed in on that browser, whether you closed or finished the getting-started guide, which Career Memory cards and groups you folded, and which empty sections you removed. These have no expiry date, and the ones kept for each account are stored under that account's ID, so they stay in the browser after you sign out or delete your account, until you change them or clear the site's data. If you arrive through a link carrying campaign tags or from another website, Radpul also keeps those tags, that website's name, the first page you opened and the time you arrived. When you sign up, everything but the time is sent with your sign-up and all of it is removed from the browser. If you do not, it stops being used after 30 days and is removed the next time you open the site. There is no advertising and no tracking across sites. Radpul measures how it is used in three ways. The first is its own count of which channels bring sign-ups and how far accounts get, from the sign-up source described above and the records described in section 3. The second is Cloudflare, which serves the website and adds its Web Analytics script to each page. When a page loads, the script reports to Cloudflare the page's address, the site that linked to it, your browser, operating system and device type, the country your connection comes from, and how long the page took to load. It sets no cookie and stores nothing in your browser, and Cloudflare states that it does not fingerprint or follow individual visitors. It counts page views, not people. The third is PostHog, a product analytics service Radpul runs in PostHog's European cloud in Germany, through an address on radpul.app. On your first visit Radpul asks whether you accept product analytics and session replay, with accepting and rejecting offered equally and nothing chosen in advance. You can change your answer at any time from Cookie preferences, at the foot of the public pages or in the account menu once you are signed in. If you accept product analytics, PostHog sets a first-party cookie named ph_phc_x7SVUBjch9BkoqqXzAEMYhuMrrtPs8MrnsEdHZbEnEHW_posthog, which expires one year after your last visit, and keeps the same identifier in your browser's local storage under the same name, so that it recognises your browser on later visits. It also keeps a few entries in the tab's session storage, which go when you close the tab, and Radpul keeps a second copy of your answer for PostHog in local storage. All of these are removed when you withdraw your consent, and the next time you open Radpul after your answer has expired. It records the pages you open (without record IDs or query, campaign tags apart), the steps you complete in Radpul, where you click and scroll, clicks that have no effect, and how fast pages load; the text and attributes of what you click are hidden before anything is sent. Once you are signed in, these carry your Radpul account ID, never your name or email address. Your IP address is used to work out where the visit came from (the continent, country, region, city, an approximate postal code and map position, and the time zone) and is then discarded. If you also accept session replay, PostHog records how you move through each page: the layout, your mouse movements, clicks and scrolling. All text and everything you type is hidden in your browser before anything is sent, and images, your CV and photo included, are not recorded, but a replay still shows the shape of each page and what you did on it. Replays are kept for 30 days. If you reject, or until you answer, PostHog sets no cookie, stores nothing in your browser except your answer, records no replay and links nothing to your account. It still counts page views and steps, with the site that linked you to Radpul: its server derives a code from your IP address and browser that changes every day, so it can tell visits apart within one day without recognising you on another, and it adds no location. The operator looks at this data in PostHog's own dashboard to see which steps work and where people get stuck. There, the events and replays of a browser that accepted analytics can be opened one by one, under that browser's identifier and, once you are signed in, your account ID. Radpul's own admin pages read nothing from PostHog. PostHog keeps events for one year and then deletes them. When you delete your account, Radpul asks PostHog to delete what it holds under your account ID, replays included; PostHog deletes the recordings, and the events in its weekly deletion run, usually within two weeks of the request. Counts made without cookies carry no account ID, so they stay until their year is over.

11. Children

Radpul is not intended for people under 18 and no account should be created for one.

12. Changes

If this policy changes, the date above changes with it. For a change that materially affects you, notice goes to your registered email address before it takes effect.

Back to home
Radpul

Built from your approved experience · AI-assisted

Radpul trains no AI model on your career data, and under the Claude API terms Anthropic does not train on it either. Admin tools never show your CV or Career Memory, and the operator does not open them unless you ask for help.