Case study · Career tech
ATS Resume Builder
Most resumes are rejected by software before a person ever reads them. This one is built to get through.
In development — join the waitlist for early access.
App screenshot
to be added
The problem
Applicant tracking systems read resumes badly
Before a recruiter sees your resume, software parses it into fields. That parser fails on exactly the design choices that make a resume look good to a human eye — and when it fails, your experience simply isn't there.
- Multi-column layouts get read across the columns, scrambling sentences into nonsense.
- Tables, text boxes, and headers/footers are frequently skipped entirely.
- Icons and graphics carrying information — contact details, skill ratings — parse as nothing at all.
How it works
Five steps, content first
- 1
Input
Paste text, upload .docx / .pdf / .txt, or import a LinkedIn “Save to PDF” export.
- 2
Category
Pick one of six professional categories. This sets templates, keyword checks, and tone.
- 3
ATS scan
A scored readiness report with severity-ranked, actionable flags — editable inline.
- 4
Template
Choose from ATS-safe single-column layouts, previewed with your own content.
- 5
Export
Download PDF and Word versions of your resume, cover letter, and references.
Under the hood
What it actually does
Every item below is shipped behavior, not a roadmap. Most of the work is in handling the messy files people really have.
Parsing that survives real-world files
Resumes arrive as whatever the user has on hand. The parser is built for that, not for clean input.
LinkedIn export support
A dedicated import flow with step-by-step instructions, plus a purpose-built parser for LinkedIn's exact education formatting — a strict line-based pass, with a lenient fallback for text that lost its line breaks during copy/paste.
In-browser PDF and DOCX reading
Files never leave the user's machine to be read. Text is extracted client-side, in the browser.
Scanned-PDF detection
If a PDF has no selectable text, the user is told plainly and offered a path forward — instead of silently getting garbage.
Unsafe-layout handling
Resumes built on tables or multi-column layouts have their text extracted and the risky layout discarded, with an explanation of why it had to go.
Batched multi-role extraction
Long career histories are pulled in successive passes with a live “found X of ~Y roles” readout, so nothing is silently dropped.
Deterministic role counting
Date-range patterns are counted directly from the source text as a sanity check against merged or missing jobs.
Truncation repair
If an AI response is cut off mid-JSON, it's walked back to the last complete value and repaired rather than thrown away.
Cleanup you can audit
Every automatic correction is shown to the user, with its before and after.
Extraction-artifact removal
Stray “Page 1 of 5” markers and LinkedIn's repeated per-page running headers are stripped automatically.
Spacing-glitch repair
Words glued together by dropped line-wrap spaces are split back apart — with a whitelist so real compound terms like CyberArk, PowerShell, and JavaScript are never mangled.
Every fix is shown, not hidden
A reviewable list of exactly what changed, before and after, with a plain statement that nothing else was touched.
Verbatim by design
The AI structures your words. It does not rewrite them.
Your wording, preserved
Content is extracted and organized without paraphrasing, shortening, or “improving” anything. The only permitted edits are mechanical extraction fixes.
Placeholder detection
Coaching text left in a template — “insert a strong action verb here” — is detected, excluded, and flagged rather than treated as real experience.
The ATS audit
0–100 readiness score
Banded Low / Medium / High, so the number means something at a glance.
Severity-ranked, actionable flags
High, medium, and low findings each paired with a specific fix — missing contact info, unquantified achievements, inconsistent dates, thin category keywords, missing critical sections.
Local heuristic fallback
A scan always returns something useful, even if the AI call fails outright.
Live staleness tracking
Edit anything after scoring and the app tells you exactly what changed and that the score is now out of date — with one-click re-check.
Full inline editing
Every parsed field is editable in place: contact block, summary, roles, bullets, education, skills, certifications, publications.
18 ATS-safe templates
Six categories, three variants each
Academic/Research, Trades/Service/Hospitality, Healthcare/Education, Creative, Tech/Engineering, and Corporate/Executive.
Single-column, always
Standard embeddable fonts, real-text section headers, standard bullet characters, reverse-chronological order. No tables, text boxes, columns, or graphics anywhere.
Style without structural risk
Templates differ by typography, spacing, hierarchy, and a single accent color — never by structure that breaks parsing.
Category-aware section order
Certifications lead for healthcare and trades, publications for academic, projects for creative and tech.
Live previews and custom accent
Thumbnails render from your own content, not stock dummy text, and the accent color is yours to pick — with reset to the template default.
Three documents, one identity
Resume, cover letter, references
An AI-drafted cover letter tailored to a pasted job description, company, and hiring contact — fully editable — plus a standalone references page.
One consistent package
All three share the chosen template's letterhead, typography, and accent, so the whole application set reads as a single piece.
Export that works
PDF with real selectable text
Never a flattened image. US Letter or A4, print-ready.
Word-compatible export
Opens and edits cleanly in Microsoft Word for recruiters who ask for an editable copy.
Sensible file names
Auto-named FirstName_LastName_Resume, _CoverLetter, and _References.
Your resume stays on your machine
PDF and DOCX files are read entirely in your browser. Nothing is uploaded to LinkedIn, and nothing is stored on a server.
Be first to use it.
The ATS Resume Builder is in active development. Leave your email and you'll get early access when it opens.
RoberOnix built this. We can build yours.
If you have a workflow that could work like this one, that's exactly the kind of project we take on.