smartvitae
Resume-fit platform where Jev scores six factual dimensions and returns KILL, FIX or SHIP before Claude adapts a resume.
Overview
SmartVitae, built by a physician, is a career alignment and resume adaptation platform designed to avoid AI hallucinations. The author's concern is that chatbots give false optimism about job fit and can fabricate credentials, which in regulated professions such as medicine can have legal consequences. SmartVitae uses a dual-engine design: TypeSafe Jev for fast System One judgment and Claude for System Two parsing and structuring. A live demo works without registration, and documentation is in English and Brazilian Portuguese.
Jev scores factual correlation between a resume and a job on six dimensions on a 0-4 scale: profession and core background fit, backing for mandatory skills, daily routine overlap, fabrication risk (whether tailoring would require inventing facts), industry fit and depth of experience. The result is a strict verdict: KILL for a radical mismatch, which blocks resume generation and offers a career roadmap instead; FIX for transferable skills with critical gaps, where the candidate can upload evidence or declare absence; and SHIP for high factual alignment, cleared for tailored generation by Claude.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | Profession and core background fit (0-4) | 0 · 1 · 2 · 3 · 4 |
| Score | Fabrication risk: would tailoring require inventing facts? (0-4) | 0 · 1 · 2 · 3 · 4 |
| Choice | Overall verdict | KILL · FIX · SHIP |
How it works
- Parse the resume and job posting
- Jev scores six factual dimensions
- Map scores to KILL, FIX or SHIP
- Only on SHIP (or resolved FIX) does Claude tailor the resume without adding facts
Strengths
- Blocks generation when fit is implausible
- Explicit fabrication-risk check
- Free live demo
Caveats
- Thresholds for verdicts not documented in the excerpt
- Uses two hosted providers
Getting started
npm install, set ENABLE_JEV=true, TYPESAFE_API_KEY and TYPESAFE_MODEL=jev-latest plus Claude credentials, then run; or use the live demo.At a glance
- Type
- Project
- Status
- prototype
- Language
- TypeScript
- Runtime
- Web app (Node)
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Plataforma de Adequação Curricular Factual e Alinhamento de Carreira com Zero Alucinação de IA. Combinando TypeSafe JEV (Julgamento cognitivo rápido System One) e Claude 3.5 Sonnet para submeter candidaturas a testes de estresse implacáveis, auditar evidências documentais e eliminar a invenção de qualificações.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
Six fit scores from 0 to 4 and the resulting KILL, FIX or SHIP verdict.
What does Claude do?
Parses documents and, when cleared, adapts the resume without inventing facts.
Why block generation?
To avoid fabricating qualifications for clearly mismatched roles.
Related projects
Resume screening where an LLM writes the questions once per job and Jev answers them for every resume, quoting the evidence line.
ProjectproductionTypeScriptCost-, cache- and quota-aware LLM router that classifies each turn with a Jev-class model, prices every route, and checks cheap answers.
ProjectprototypePythonResume knockout tool: deterministic date math and hard gates first, then a Jev decision gate, producing disqualify or advance verdicts.
ProjectprototypeChecks whether each cited paper supports its citing sentence: code fetches papers, Claude proves a quote, Jev scores it, a human decides.
ServiceproductionPublished Jev client and claim check from the Fake/Real evidence-review site, where Jev filters sources and judges each claim.
ProjectproductionTypeScriptZilliz semantic cache for LLM queries, listed for a Jev evaluator that uses Noul checks to decide whether a cached answer can be served.
LibraryproductionPython