Role Radar: An AI-Based Resume Intelligence and Skill Gap Analysis System

Job seekers face two linked problems: finding postings that are genuinely open and suitable, and presenting their experience to automated screening without exaggeration. General-purpose language-model resume writers aggravate the second problem because they can invent skills, metrics, and seniority. This paper presents RoleRadar, a full-stack career platform built on one rule: deterministic code decides what is true, and the language model only rephrases it. A geometry-based parser turns an uploaded resume into atomic, provenance-tagged evidence units. Job requirements are mapped to that evidence on an eight-level support scale, and a category-aware six-factor model yields an explainable match score and a readiness verdict. A rule-based Truth Guard compares every tailored bullet with its source and reverts fabricated technologies, unsupported metrics, inflated titles, and broken sentences before one-page export. Supporting modules provide a taxonomy of 119 roles, prerequisite-ordered learning roadmaps, verified job ingestion from direct employer feeds, and an assistant restricted to the signed-in user’s data. A regression suite of 1,268 test cases, internal validation harnesses, and an end-to-end acceptance run indicate that the pipeline is consistent and auditable. The system has not yet been evaluated with recruiters or in a controlled user study.