resurface
Resume screening where an LLM writes the questions once per job and Jev answers them for every resume, quoting the evidence line.
Overview
Resurface turns an existing candidate database into evidence-backed shortlists. An LLM writes screening questions once per job description; then TypeSafe Jev answers them for every resume in one request each, and also picks the exact resume line that proves each answer via a Choice over line ids. A plain-language pool search uses a Noul (does this resume fit the request?) plus a line choice for justification. The list summary reports it as 8x faster and 31x cheaper than an LLM, and the repo includes a head-to-head benchmark script against Claude models on latency, cost, consistency and agreement.
This repository is the MIT-licensed open-source version: clone it and add TypeSafe and LLM keys. A hosted Resurface Cloud API returns a 0-100 score, a verdict per question and the proving line, with 50 free scores then $0.10 per resume, and is available as an MCP server, a Claude plugin and an agent skill. The open version has basic PDF import and ephemeral storage.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does this resume meet screening question X? | — |
| Choice | Which resume line proves this answer? | — |
| Yes / no | Does this resume fit the request? | — |
How it works
- LLM drafts screening questions from the job description
- For each resume, one Jev request answers every question and picks evidence lines
- Combine answers into a score and shortlist
- Show each verdict with its quoted evidence line
Strengths
- Evidence line quoted for each verdict
- List summary reports 8x faster and 31x cheaper than an LLM
- One request per resume
Caveats
- Open version has basic PDF parsing and ephemeral storage
- Hiring decisions need human review and fairness checks
Getting started
Clone the repo and add TypeSafe and LLM keys, or use the hosted API (curl https://getresurface.dev/api/v1/score) or npx skills add dolevhayut/resurface.At a glance
- Type
- Project
- Status
- production
- Language
- TypeScript
- Runtime
- Web app, MCP server, Claude plugin
- Has code
- yes
- License
- MIT
- Stars
- 0
Use cases and integrations
How curated lists describe it
TypeSafe JEV resume screening: LLM writes the questions, JEV (System One classification model) answers them for every resume in your ATS with the exact evidence line. 8× faster, 31× cheaper than an LLM.
wh000wh000/awesome-jev-live
Frequently asked questions
What does Jev decide?
The answer to each screening question for each resume and the line that proves it.
What does the LLM do?
It writes the screening questions once per job; it does not score resumes.
What does the hosted version cost?
50 free scores, then $0.10 per resume per the README.
Related projects
Resume review app and MCP server where a Cloudflare Worker asks Jev typed questions and composes a JevScore, generally or per job.
AppbetaTypeScriptZero-dependency library that numbers a text's words so Jev can pick them, returning exact substrings for extraction and PII.
DemoprototypeTypeScriptLibrary, CLI and app that classify PDFs, DOCX and PPTX with Jev and split document packets at predicted boundaries.
BenchmarkbetaPythonLarge Claude Code plugin of skills, agents and hooks that can optionally ask Jev to classify a session's work category.
ProjectproductionTypeScriptClassifies tax PDF pages into 261 IRS forms and 7 page kinds with one Jev request per page and a confidence gate.
ProjectproductionTypeScriptRust tool that asks Jev which PDF pages genuinely need OCR, extracting text pages locally and sending only the rest to OCR.
ProjectbetaRust