su misura — Italian for “made to measure”
Resumes tailored to every job,
never made up.
Sumisura turns your complete career history into a one-page CV and a cover letter written for one specific job — selected from what you actually did, rewritten under a groundedness check, and rendered only after you approve it.
Self-hosted · your files, your API key · AGPL-3.0
AI resume tools have one job, and they cheat at it
Ask a generic builder to tailor your CV to a posting and it will happily give you Kubernetes experience you don't have, invent a percentage to make a bullet sound sharper, or quietly promote you a level. You find out in the interview.
Sumisura is built the other way around. Rewriting is allowed to changeemphasis and wording. It is not allowed to add a fact. Every rewritten bullet is checked against the source bullet it came from, and the check runs before you are ever asked to approve anything.
How a tailored CV gets made
- 01
Keep your history once
Every job, client engagement and project lives as a Markdown file with frontmatter — your Master Data. Write it once; it is the only thing you maintain.
- 02
Point it at a job
Paste a job description or URL, pull it from a Greenhouse/Lever/Ashby board, or capture the LinkedIn or Indeed posting you are reading with the browser extension.
- 03
Selection, then Rewrite
Claude picks the Entries that matter for this job and rewrites their bullets towards its language — never inventing a new fact, tool or number.
- 04
You approve, then it renders
Text Review and Visual Review are real checkpoints. A one-page PDF and a matching cover letter come out of a Typst template you control.
The groundedness check, in one example
Built the ingestion service that feeds the recommendation models, moving it from nightly batches to streaming.
Rebuilt a batch ingestion pipeline as a streaming service feeding production recommendation models.
Same facts, the job's vocabulary.
Cut recommendation latency by 40% for 12M users usingKafka and Flink.
Numbers and tools that appear nowhere in your history.
It also keeps track of where you applied
Every job you save becomes a Job Listing with an Application attached: status, timestamped notes, the contact, the salary range where it can be resolved, and every Generation you produced for it. A browser extension captures the LinkedIn or Indeed posting you are looking at; ATS boards (Greenhouse, Lever, Ashby) can be pulled directly.
Self-hosted, and private by construction
- Runs on your machine via
docker compose, bound to localhost. - No accounts, no sign-in, no telemetry, no phone-home.
- Your Master Data is plain Markdown and YAML in your own directory.
- It calls the Claude API with your key; nothing goes anywhere else.
- Your output is yours — the AGPL puts no claim on what you generate.
Would you use a hosted version?
Self-hosting asks for Docker and an API key. A hosted Sumisura would not. Before building it, it is worth knowing whether anyone wants it — so there is no waitlist and no email box here, just a thread.
👍 the discussion on GitHub