Career Command & TypeSafe Jev (jev-latest) Eval Engine
Full-stack Astro + Cloudflare D1 intelligence platform combining 22 automated ATS ingestion scanners with 6 deterministic TypeSafe Jev primitives for role qualification.
STAR Engineering Breakdown
01 · Situation & Problem
Keyword and regex filters across job boards produced high false-positive rates, misclassifying hybrid roles as remote and failing to distinguish Staff Technical SEO roles from generic content marketing roles.
02 · Technical Execution
Architected an Astro + TypeScript platform backed by Cloudflare D1 that ingests postings from 22 ATS endpoints and evaluates every listing through TypeSafe System One (jev-latest) typed primitives.
03 · Composable AI Primitives
Chained jev.boolean for hard eligibility gates, jev.choice for primary role-track routing, jev.multichoice for capability tagging, and jev.score for calibrated weighted fit scoring without brittle LLM JSON parsing.
04 · Verified Impact
Eliminated brittle regex heuristics in favor of deterministic, probability-backed TypeScript judgments that automatically match each role to verified resume proof points and case study blueprints.
const isRemoteEligible = await jev.boolean(
posting.locationText,
"The role allows full-time remote work from the United States"
);
const roleTrack = await jev.choice(posting.summary, ROLE_TRACKS);
const fitScore = await jev.score(posting.requirements, RUBRIC_SPEC);Benchmark Table
| METRIC | MANUAL BASELINE | PROGRAMMATIC ENGINE |
|---|---|---|
| Role Qualification Accuracy | Brittle Keyword / Regex Rules | 6 Typed Jev Semantic Primitives |
| ATS Ingestion Coverage | Manual Board Checks | 22 Automated ATS API Scanners |
| LLM Output Reliability | Unstructured Prompt-and-Parse | 100% Type-Safe Probability Primitives |