RoleRamp: AI CV tailoring built as a reusable product base
A career toolbox that tailors a CV and cover letter to any job ad from a single base CV, exports ATS-friendly documents, and tracks every application, later reused as the base SnapKai and Lernoa were forked from.
Visit RoleRampProblem
Rewriting a CV and cover letter for every job application is slow and repetitive, and most job seekers either skip tailoring or do it badly under time pressure. RoleRamp needed to turn one base CV into a genuinely tailored, ATS-friendly application for any job ad, while tracking every application in one place.
Architecture
Frontend
A React/Redux single-page application for the product, plus an Astro marketing site and a browser extension.
API
Two .NET 10 APIs backed by Postgres, plus an internal admin application.
Database
PostgreSQL.
Infrastructure
AWS and Docker, deployed via CI/CD to the shared Auckland EC2 infrastructure used across the Corvidae fleet.
AI components
LLM extraction of CV content from uploaded PDF or DOCX files, per-job-ad LLM tailoring of the CV and cover letter, AI-assisted rewriting for quantified impact, and an ATS-friendly PDF/HTML export pipeline running on AWS Bedrock and Textract.
Engineering decisions
- The base CV is kept as a single source of truth, with every tailored version generated from it rather than manually duplicated and left to drift out of sync.
- Document export runs through a WeasyPrint-based pipeline built into the .NET API, producing ATS-friendly PDF and HTML output rather than relying on a client-side renderer.
- Billing is multi-currency (NZD, AUD, USD, CAD, GBP, EUR) one-off "Role Pack" credits rather than a single-currency subscription, since the audience spans several countries.
- Because RoleRamp became the base other Corvidae products were later forked from, its authentication, billing, admin and credit systems were deliberately built to generalise beyond a single product rather than being hard-coded to CVs.
Challenges
- Extracting structured CV content reliably from arbitrary uploaded PDF and DOCX formats with inconsistent layouts.
- Generating tailored output that reads as genuinely rewritten for the job ad rather than a templated substitution.
- Building a public MCP server on top of the existing product, including its own OAuth 2.1 authentication and a tool-permission guard, without disrupting the API the web app already depended on.
Outcome
- In production with Stripe live keys, webhook handling, OAuth and the document export pipeline all verified working.
- Live at roleramp.com with a companion browser extension for capturing job ads.
- A public MCP server at api.roleramp.com/mcp exposes dozens of tools under OAuth 2.1 to AI assistants.
- Became the reference production base that SnapKai and Lernoa were later forked from.
Services demonstrated
Other case studies
Want something built this way?
Tell us what you want to build. We reply within one business day with an honest read on scope, cost and timeline.
Start the conversation