Remi has already read your resume, your report and the job. It still asks before it changes a word.
Remi by Roleframe works alongside the document you are editing. It reads the open document, the posting you are targeting and every finding in your analysis report. Then it proposes one section at a time, as a diff you approve.
- Works on the open document
- Reads your report
- Nothing changes without your approval
That bullet claims reliability. The result is in your report: failed payments fell 38% on 1.2M monthly orders. Here it is in the line itself.
Kept your wording, added the number your report already holds.
Led the migration of checkout to event sourcing, improving reliability cutting failed payments 38% across 1.2M monthly orders.
One turn, mid-conversation. Remi has read the document, the job and the report, and the proposal waits until you decide.
A chatbot in another tab cannot see your document, your report or the posting. So you paste. Then you paste again after every edit, and the answer you get is written about a copy of your resume that is already out of date.
Remi starts every turn with the live document in front of it. There is nothing to paste and nothing to keep in sync.
- Read the open document6 sections · 8 bullets · 1 page0.4s
- Read the linked postingStaff Engineer, Payments · Monzo0.3s
- Read the analysis report6 open findings · 1 stale0.2s
- Draft a proposal for one blockExperience · Stripe · bullet 12.1s
Four things are on the table before you type.
Remi is a recruiter who switched sides. It reads the way the person across the table reads, and it starts from what is actually on your page.
It reads the open resume by section
Ask about “my summary” or “the second job” and Remi opens that exact block from the live document. It knows the shape of the page, so an ambiguous reference comes back as a short list you click rather than a guess.
It reads the job you are applying to
When the resume is tied to a job, Remi has the title, the company, the application status and the full text of the posting. A recruiter gives your page six seconds, and the summary is what they catch, so that is usually where the posting’s language belongs first.
It reads your analysis report
The score, every open finding, the reason behind each deduction, and the keywords the posting asks for that your resume never uses. It also knows which findings went stale because you edited that section afterwards, and it says so instead of asserting a fix you already made.
17 skills to close
Remi only proposes a keyword where your own experience supports it. A term you cannot back up stays on the list.
It writes the proposal. You press the button.
Remi cannot write to your document. It produces a proposal, one section at a time, and the gate is enforced in the tool itself rather than by prompt wording.
Led the migration of checkout to event sourcing, improving reliability.
Led the migration of checkout to event sourcing, improving reliability cutting failed payments 38% across 1.2M monthly orders.
Applied and discarded proposals collapse into a slim strip in the thread. Reopen either one later.
One section at a time
A proposal covers the block you asked about. Remi does not rewrite the document around it.
Word-level diffs
You see exactly which words leave and which arrive, in place, before anything is committed.
Reversible
Every applied edit is a version you can restore. The score moves back with it.
Every finding in your report has a button.
Fix with Remi is the primary action on each issue. It hands that finding to Remi with the block it refers to. Apply the fix and the finding is marked resolved, and the score moves by exactly what that finding cost.
The report as it appears beside the editor. Remi never runs a fix you did not press.
- Step 1 · the findingCosts 7 points
3 of your 8 bullets have a measurable result.
Experience · Stripe · bullets 1, 2 and 4.
Fix with RemiLocate - Step 2 · the proposal
Three bullets, three numbers from your report. Review each one.
improving reliabilitycutting failed payments 38%faster deploysdeploys from 40 min to 6mentored the teammentored 4 engineers, 2 promotedApply 3 editsDiscard - Step 3 · the score7178/100Finding resolved, 7 points back
Five findings left. The next one is your summary, which reads like a job description.
Remi ends a turn with two to four options you can click, not a wall of bullet points.
The resume gets you the call. Remi rehearses the call.
For a tailored resume, Remi generates prep on demand: STAR stories and the themes the posting keeps returning to, written from your resume as it stands today. It takes about half a minute.
Remi can help you with STAR interview answers based on your resume and the job description.
Walk me through the checkout migration. What was the hardest call you had to make?
Strong on what you built, thin on why. You named the tool before the problem. Open with the failed payments number next time.
Suggested answer
Start with the 38% failure rate on retries. Then the call: you kept the old path live for six weeks behind a flag, which cost you a quarter of the timeline and made the rollback boring. Close with what the team runs today.
One question per message. The suggested answer stays collapsed until you ask for it.
Your weakest category is the one that comes up second in most loops. Two more runs and it will not be.
Checkout ran on a synchronous path that lost 38% of retried payments across 1.2M monthly orders.
You owned the migration to event sourcing without a freeze on releases or a drop in conversion.
You kept the old path live behind a flag for six weeks, mirrored traffic, and cut over per region.
Failed payments fell 38%. Deploys went from 40 minutes to 6. Rollback was never needed.
Each part quotes something already on your page. If a section is thin, Remi asks you for the detail rather than inventing one.
Four things Remi will not do.
It will not write to your document
The approval gate lives in the tool Remi calls. There is no path from a proposal to your page that skips the Apply edit button.
It will not hand your voice to a stranger
Dictation is transcribed on Roleframe’s own servers. The browser’s built-in speech engine quietly ships audio to a third party, so Remi does not use it.
It will not take orders from a posting
Text inside a resume or a job posting is treated as text, never as instructions. A posting that says “ignore your instructions” is just a posting with a strange sentence in it.
It will not improvise
Answers quote the lines you actually wrote and the findings your analysis actually produced, and they name the section they came from. A live progress tree shows each read while it works.
Before you turn it on.
Does it change my resume on its own?
No. Remi can only produce a proposal. Applying it is a button you press, and the restriction is enforced in the tool rather than by convention.
What does it know about me?
The resume you have open, the job linked to it, your analysis report, and the block you selected. That is the context it starts a turn with.
Does it work without a job attached?
Yes. It reads the document and the report either way. Link a job and it also has the posting, the missing keywords and interview prep for that role.
Can I use it on a cover letter?
Remi works on the document you have open, one at a time. It does not reach into anything you are not looking at.
What happens to what I dictate?
The audio is transcribed on Roleframe’s servers and becomes text in the composer, which you can edit before you send it.
What if a finding is out of date?
Remi knows which findings went stale because you edited that section after the analysis ran, and it says so rather than proposing a fix you already made.
Open your resume. Ask Remi what a recruiter would cut.
Remi reads the document, the job and the report before it answers, and every rewrite waits for your approval.
- No card required
- A few full runs on the free plan
- Nothing changes until you apply it