Recruitment Assistant Agent

Agent Description:

The Recruitment Assistant Agent is designed to automate the initial stages of candidate screening by transforming unstructured CVs into actionable intelligence. It enables HR teams to objectively evaluate large volumes of applicants by extracting key professional attributes, identifying skill gaps, and calculating holistic "fit" scores based on weighted organizational requirements.

Purpose and Components
  • Purpose: This template streamlines the recruitment funnel by parsing various document formats (PDF, DOCX, Text), validating the clarity and completeness of extracted data, and dynamically routing profiles for either specialized profile matching or direct skill evaluation. The agent eliminates manual data entry and reduces screening bias by providing a standardized, multi-dimensional scoring rubric for every candidate.

    The agent improves talent acquisition efficiency by:

    • Extracting and normalizing core CV fields (Skills, Experience, Education) with associated confidence scoring.

    • Holistically evaluating CV quality to flag low-detail or unstructured submissions.

    • Automatically comparing candidate skill sets against specific job descriptions using semantic similarity.

    • Computing weighted fit scores across four pillars: Skills, Experience, Profile Consistency, and Engagement.

    • Categorizing candidates into clear labels (Excellent, Moderate, or Low Fit) with detailed justifications.

  • Components:

    • CV Extractor: Converts unstructured resumes into structured JSON and assigns a confidence score based on data completeness.

    • Profile Evaluator: Acts as the primary workflow router, directing high-confidence data to skill evaluators and low-confidence data to profile matchers.

    • Profile Matcher (A2A Gateway): An external gateway used to enrich or reconcile low-confidence candidate data.

    • Skill Evaluator: Performs a deep-dive comparison between applicant skills and the target job description to compute a match score.

    • Final Scorer: The aggregation engine that calculates a weighted candidate profile and produces the final fit summary.

Supported Capabilities

  • Multi-format CV parsing (PDF/DOCX/Text)

  • Holistic confidence scoring (0.0–1.0) based on section completeness

  • Data normalization (Name, Skills, Experience, Education)

  • Semantic skill similarity analysis and keyword matching

  • Missing vs. Strong skill identification

  • Weighted scoring logic: 40% Skills, 30% Experience, 20% Consistency, 10% Engagement

  • Conditional workflow branching based on confidenceScore

  • Automated candidate labeling (Excellent, Moderate, Low Fit)

LLM Used

  • OPENAI GPT_4O_MINI

    Note: To learn more about the LLM and to modify its behavior, refer to the Configuring LLM settings section.

Sub-Agents

1. CV Extractor

  • Role:CV Extractor

  • Scope:Converts unstructured CVs into structured JSON with confidence scoring.

  • Description:Extracts name, skills, experience, and education. It evaluates the CV's clarity holistically (for example, presence of measurable results) to assign a score where >0.75 is "Good" and <0.4 is "Very Weak."

  • LLM Used: Default (Inherits from parent agent).

2. Profile Evaluator

  • Role:Profile Evaluator

  • Scope:Evaluates data quality and selects the next node for handoff.

  • Description:Acts as the logic controller. If structured_data.confidence < 0.7, it routes to the Profile Matcher; if > 0.7, it routes directly to the Skill Evaluator for technical assessment.

  • LLM Used: Default (Inherits from parent agent).

3. Skill Evaluator

  • Role:Skill Evaluator

  • Scope:Evaluate candidate skills and experience against specific job requirements.

  • Description:Uses keyword similarity and embeddings to identify missing vs. strong skills. Computes a match score (0–1) and flags any profile scoring below 0.75.

  • LLM Used: Default (Inherits from parent agent).

4. Final Scorer

  • Role:Final Scorer

  • Scope:Generate a holistic, weighted evaluation of the candidate.

  • Description:Aggregates all data to compute a final weighted score (Skills, Experience, Consistency, Engagement). Produces a summary labeling the candidate as an Excellent, Moderate, or Low Fit.

  • LLM Used: Default (Inherits from parent agent).

Tools Used:

  • A2A Gateway- Connects to the external profile-matcher agent to handle low-confidence data enrichment.

Note: For details on modifying the Tools, refer Tools Library section.
Agent Workflow Behavior Summary

  1. Input: A CV file is uploaded to the CV Extractor.

  2. Validation: The Extractor identifies fields like "Python" and "5 years experience." If the text is messy, it assigns a confidenceScore of 0.6.

  3. Routing: The Profile Evaluator sees the 0.6 score and sends it to the Profile Matcher (A2A) for reconciliation. If the score was 0.9, it would skip to the Skill Evaluator.

  4. Scoring: The Final Scorer receives the skill match (for example, 85%) and experience relevance (for example, 70%) to calculate the final percentage.

  5. Final Output: A comprehensive candidate report with a fit label and justification.

Sample Questions:

  • Screen the attached CV for the Senior DevOps Engineer role.

  • Evaluate candidate John Doe's profile against our mandatory skill list for Project Managers.