Back to work

Case study / MULTI-AGENT SYSTEM

RecruitRadar

A CrewAI multi-agent recruitment screener that scores resumes against a job description with deterministic, evidence-backed scoring — saving recruiters hours per role.

PythonCrewAIGeminiLiteLLMrapidfuzzSQLModelSQLite
View on GitHub

Architecture

JD AnalystResume ScorerInterview DesignerOutreach DrafterTrackerEvidence guardrapidfuzz · every claim cites the resume

How it works

01

JD Analyst

Parses the job description into weighted, structured criteria that drive every downstream decision.

02

Embedding pre-filter

A Gemini embedding pass shortlists resumes so expensive scoring only runs on relevant candidates.

03

Resume Scorer

Computes a deterministic 0–100 score with weighted sums in code (never emitted by the LLM); every matched skill must cite a verbatim resume quote, verified with rapidfuzz.

04

Interview Designer

Drafts tailored interview questions grounded in the candidate’s actual experience and the role’s criteria.

05

Outreach & Tracker

Writes personalized outreach messages and tracks every candidate through an SQLite kanban pipeline.

Engineering highlights

~6 hours saved per role screened

Deterministic, auditable scoring — no LLM-emitted numbers

Evidence guard: every claim cites a verbatim resume quote

Next project

SmartIngest