Privacy Policy

Version 0.2 · October 2026

This describes what PrepHike holds about you, why it holds it, who else gets to see it, how long it stays, and how you destroy it. It is written against what the software actually does rather than against what would sound best. Where the mechanism is weaker than you might hope, this page says the weaker thing. Version 0.2 exists because version 0.1 failed that test in several places, and each correction is marked where it sits rather than quietly swapped. It sits alongside the terms.

01 What this covers

Two surfaces. This website, prephike.com, which is informational and processes no document you supply. And the application at app.prephike.com, which needs an account.

02 What you give us, and why

Your resume. You give it to us after you sign in. There is no check you can run without an account. Upload a PDF and the file is read inside your browser, never leaving it. Only the text pulled out crosses the network. Paste text instead and that text crosses the network. It is stored as ordinary text in an ordinary column, exactly as you gave it. Every finding quotes it at character offsets, and those offsets have to keep pointing at the same characters. It is not encrypted. An earlier version of this page said it was. That was wrong, and section 05 now describes what actually protects it.
The role you are aiming for. Free text you type, stored as an ordinary column. Your resume is read against it.
Your connecting IP address, briefly. The free check counts requests against the address your connection arrives from, so one machine cannot run the check thousands of times. The counter is keyed on the address, lives in a short lived store, and expires with its window. It is not written next to your check and it is not kept.
Your email address. It is stored, in full and readable, on your account record. It has to be. It is what your account is keyed on, and where a sign in link goes. The application shows it so you know which account you are in. Signing in with Google also returns your name and your profile picture, and both are kept if Google supplies them. Everything in this bullet is deleted when you delete your account.

Until 2026-10-25 this page said the opposite: no email column, only a hash. That was false, and false from the first version. One thing is hashed, the sign in link token, so a used or stolen link cannot be replayed. The address beside it is plain text.
What you type in a mock interview. Your answers are stored as ordinary text, up to 8,000 characters per turn, so the interviewer can answer and so you can read the session back. See section 05.
Your offer figures, if you enter them. What you earn now, what you are aiming at, and how long your notice is. They are stored as ordinary whole numbers in ordinary columns, so a query could in principle filter or average them. Nothing we publish counts them, and they are never sent to any model. An earlier version of this page claimed they were encrypted and that no column existed to filter on. Both were wrong.
The companies you are talking to, if you record them. The company name, the role title, where it came from, the stage it reached, the rounds you log and how each ended. All of it is ordinary text and numbers. A company name is a fact about somebody other than you, which is why it is called out here. An earlier version said the name and role were encrypted. They are not.
A payment, if you buy something. What is stored on our side is an internal order id, a product code, an amount in paise, a currency, a status, and the payment provider's own reference. No name, no email address, no card number and no card fragment. Card details are entered with the payment provider and never reach us.

03 What we work out, rather than ask for

Findings. Each one quotes a span of your own document word for word, recorded with the character offsets the quote was taken at. Anything that cannot be found verbatim in your text is discarded before you see it.
A record of every model call made for your check: which model, which provider, how long it took, what the gates decided, and a hash of the exact request that was sent. The request itself is never stored, because it contains your resume. The model's raw reply is stored, and because the model is required to quote you, that reply contains sentences copied out of your resume. It is kept so a disputed finding can be reconstructed exactly. It is deleted when you delete your account.
Counts and timings from your use of the application, used to show you where you are and to work out whether the product helps anybody.

04 What we do not have

We do not have your phone number, your postal address or your date of birth. Nothing asks for any of them.
We do not have a password, because this product has none anywhere. You sign in with a link or with Google.
We do not have your card number or any fragment of it. Card details are entered with the payment provider and never reach us.
We do not have your current employer's name unless you type it into your resume or record it as a company you are talking to.
We do not have an advertising profile of you, anywhere, by any name.

This list used to be longer. It claimed we held no name, no email address and no photograph. All three were wrong: your address is on your account record, and Google sign in supplies a name and a picture which are kept. They are listed in section 02 where they belong, and they are deleted with your account.

05 Where it lives, and who runs the machines

Everything runs on Cloudflare. The code runs as a Worker, the database is Cloudflare D1, and this website is served by Cloudflare Pages. The rate limit counter is a Workers rate limiting binding. The website keeps one short lived secret in Cloudflare KV, which is the 24 hour salt described in section 08. Cloudflare is therefore the infrastructure provider for all of it, and where data physically sits is a function of their network rather than a choice we make per request.

Now the part this page got wrong, stated plainly, because it is the most important thing on it.

Your text is not encrypted by us. Until 2026-10-25 this page called your resume, your offer figures and the company names AES-GCM ciphertext, held under a key we could destroy. No such encryption exists in the software. None ever did. It was designed, written down here, and never built. Your resume sits in a database column as the text you pasted. So do the company and role names. Your offer figures sit in number columns.

There is a reason it cannot simply be switched on where the resume is concerned, and you are entitled to it. Every finding this product shows you quotes your document at exact character positions, and that is the promise the whole thing rests on. Character positions into ciphertext point at nothing. A product that encrypts the resume at rest is a different product from one that quotes it back word for word. We built the second.

What does protect it. The database is not reachable from the internet, only through code that resolves who you are first. Every route that returns your text proves ownership against your session before it answers. Nothing goes to anyone section 07 does not list.

What does not protect it: a key, because there is no key. If someone read Cloudflare's database who should not have, your resume would be readable. We would rather write that than leave the earlier sentence standing.

What you type in a mock interview is the same: ordinary text in an ordinary column until you delete your account, at which point the rows are deleted outright.

06 Who else sees it

The model provider. To read your resume against a role we send that text and the role you named to OpenRouter, which routes it to the pinned model we use for that step. The same applies to a debrief you ask us to read, and to a mock interview, where what you type is sent so the interviewer can answer. Your offer figures are never sent. Your recorded company names are never sent.
Cloudflare. As the infrastructure described in section 05. There is no separate anti bot challenge on any form; an earlier version of this page said there was.
Whoever your browser uses to turn speech into text. A spoken mock interview and a spoken debrief are transcribed by your own browser, not by us. That is not the same as staying on your device. Chrome sends the audio to Google, Edge sends it to Microsoft, and Safari handles it on Apple's terms. We never receive the audio and we never store it. What reaches us is the text, saved against the round the way typed text is.
Nobody, if you type. Typing sends no audio anywhere. Firefox has no speech recognition, so on Firefox typing is the only route and the spoken path is simply unavailable.
Google. Only if you choose Google to sign in, in which case Google knows you signed in to PrepHike. Separately, both this website and the application load their fonts from Google's font servers, so Google's servers see the request your browser makes for a font file.
The payment provider, Razorpay. If you buy something, you transact with them. They see what a payment needs. We see their reference for it.
An email provider, Resend. It delivers sign in links, which means an address you type for a sign in link is handed to it. This page named Brevo until 2026-10-24, which was wrong: no part of the product calls Brevo, and none of it ever did. Read section 14 before assuming mail works at all.

07 Who never sees it

We do not sell your data, and there is no arrangement under which anybody pays us for it.
We do not send your resume, your figures or your search to recruiters, employers or the companies you are talking to. We are not a placement agency and we do not contact companies on your behalf.
There is no advertising network, no advertising pixel, no data broker and no third party session recorder anywhere on either surface.
Anything we publish about how well the product works is counted from records that carry no name, no email address and no resume text. Your figures are never counted into it, because there is no query that could count them.

08 Measuring this website

This website records what pages were opened, what was clicked, roughly what size of screen it happened on, which page sent you here, and the country Cloudflare's edge reports. It writes those to Cloudflare's analytics store. It stores nothing at all on your device: no cookie, no local storage, nothing to clear. It does not store your IP address.

To count how many different people visited on a given day, our server works out a short code for your visit. It is made by taking a random secret, your IP address, your browser's user agent string and the website address, and running them through a one way hash. Only that code is kept. Your IP address and user agent are used for the length of one request and then dropped.

The random secret is thrown away and replaced every 24 hours, and we keep no copy of the old one. So the code you get today cannot be recreated tomorrow, and yesterday's visit cannot be joined to today's, by us or by anybody else. The price of that is a real one and we would rather say it than hide it: we cannot tell a returning visitor from a new one, and we do not try.

Two people sharing one internet connection with the same browser can land on the same code, and one person whose connection changes during the day can produce two. The number is a rough count of visits, not a count of people, and nothing we publish treats it as one.

09 How long we keep things

Your resume, your applications, your rounds and your debriefs: kept until you delete them or delete your account. There is nothing you can run without signing in, so there is no anonymous record to expire. Earlier versions of this page described a free check you never sign in for, and a shareable result link that expired after 30 days. Neither exists in the software.
A record of what you agreed to: kept until you delete your account. Two things are recorded, each once: that you accepted these terms and this notice, and that you uploaded a resume. Each row holds the purpose, the version of this page you were shown, and the date. So which notice you saw, and when, is answerable rather than assumed. Nothing else is recorded. Analytics keep no identifier that outlasts the day, and nothing here sends you marketing email.
Rate limit counters: minutes. They expire with their window.
The record that a reading happened: kept indefinitely, including after you delete your account, and the database physically refuses to delete it. It carries no name, no address and no text from your documents. What it contains is in section 10. It holds no score, because this product does not put a number on you.

10 Deleting your account

You can delete your account from inside the application. It asks you to type a confirmation phrase, because there is no backup, no undo, and no support process that can reverse it. What runs is nine passes, in this order.

The mock interview sessions and their transcripts.
The debrief findings and transcript segments.
The debriefs themselves.
The competency observations.
The job descriptions and their requirements.
The applications with their rounds and recruiter notes.
The payment records and credits.
Your name on the operator record described below.
Last, your account with your email address, your sign in methods and the record of what you agreed to.

Your resume, your offer figures and your profile go with those passes. Your browser session is cleared as it finishes, so the address cannot sign in afterwards because there is nothing left to sign in to.

The application states the outcome in these words, and this page repeats them rather than paraphrasing them:

Deleting your account deletes your resume text and your account details outright. The rows are removed, not hidden and not marked. There is no undo and no support process that can bring them back.
The record that a reading happened stays. It carries no name, no email address and no text from any document you supplied: which surface asked for the reading, which model answered, how many findings survived grounding, and timestamps, keyed to a random identifier that after deletion points at nothing. It stays because it is how we measure, across everyone who has used this, whether the product actually helps.
If you record a company you are talking to, we keep it: the company, the role, the stage it has reached, the rounds you log against it and how each one ended. It is never shown to anybody else and it is never counted into anything we publish. Deleting your account deletes those rows, so the record of who you were talking to does not survive it.
Putting an application away is not deleting it, and the screen says so where you do it. As of 2026-10-25 you can take one off your list, with its rounds. The row stays, marked with the date you put it away, and the panel it moves to lets you take it back. It is kept rather than destroyed because your write ups against it, and the skills read out of those, cite it. Destroying it would move your own numbers months after you earned them. Deleting your account deletes it with everything else.

One record survives on purpose, and it is new as of 2026-10-24. An operator may have acted on your account, issuing a refund or granting credits. The record that the action happened stays: what was done, which operator did it, and when. Your identifier is removed from it, so it names nobody. It stays because it is what makes money reconcilable after an account is gone, and it carries no document text and no address.

The word here is deleted, and it used to be unreadable. Until 2026-10-25 this section said your resume survived deletion as encrypted bytes whose key we threw away. There are no such bytes and there is no such key. What happens is simpler and, as it turns out, stronger: the rows are deleted. A row that would still say something about you even emptied of its text goes too. An offer note recording that you had an offer on a date is removed, not blanked. So is an application recording that you reached round three somewhere and stopped on the eleventh.

And the sentence the application refuses to leave out: this is not a certified erasure under any named regulation and it has not been reviewed by a lawyer. It is what the mechanism above does, described as plainly as we can put it.

11 Taking your data out

The application will give you one JSON file containing what it holds about you: your account record, every reading with its target role, the text of the document you gave it, and the findings with the spans they quote. There is no score in it, because the product does not produce one. Where something is missing the file says why rather than quietly leaving it out. You do not have to be leaving to ask for it.

One thing is deliberately left out of the export: the raw model reply described in section 03. It is an internal audit artifact, and the evidence span it contains is already on every finding in the file. Leaving it out of the export changes nothing about deletion, which removes it either way.

The export takes no identifier of any kind. It returns what belongs to whoever is signed in, and there is no link, no token and no administrative route that reaches it.

12 Age

PrepHike is for working software engineers. It is not intended for anyone under 18, and we do not knowingly collect anything from anyone under 18. If you are under 18, a parent or guardian must agree on your behalf, as the terms say. If you believe a person under 18 has given us something, tell us using section 14 and we will destroy it.

13 Changes to this policy

We may update this over time, and we intend to update it whenever the software starts holding something it did not hold before, rather than afterwards. The version and date at the top of this page say which one you are reading. The policy that applies to you is the one in force when you use the service.

14 Contacting us, and a gap we are not hiding

PrepHike is based in Coimbatore, India. Write to [email protected], which is the same address our terms carry.

Now the honest part. The software sends one kind of message and one only: your sign in link. It sends nothing about your purchases and nothing about your data, and there is no marketing list to be on. Until 2026-10-25 this page said the product could not send email at all. That was true when written. It was tested on that date and a sign in link arrived, so the claim is withdrawn. Any reply to the address above comes from a person reading that mailbox. We will not promise a response time we have not earned. We have not appointed a named grievance officer or a data protection officer. When we do, this page will say who they are.

Because of that gap, the two things you are most likely to want do not depend on reaching us at all. Your export and the deletion described in section 10 are both controls inside the application, they run immediately, and nobody at PrepHike has to be involved.