Match job descriptions with resumes using Google Gemini and log scores to Google Sheets

Smart Resume Screener — JD ↔ Resume AI Match & Sheet Logger

Smart Resume Screener ingests a candidate resume and a job description link, extracts clean text from both, runs an LLM-powered screening agent to produce a structured assessment (strengths, weaknesses, risk/reward, justification, and a 0–10 fit score), extracts contact details, and appends a single, validated row to a Google Sheet for tracking.

---

How It Works (Step-by-Step)

1. Trigger — On Form Submission

Public form webhook sends:

• Binary resume file (PDF / DOCX)
• Job Description (JD) URL or text

---

2. Extract & Fetch Content

• Resume Extraction node

• Converts the uploaded binary resume into plain text (data.resume).

• HTTP Request node

• Fetches the JD HTML/text from the provided link.

• Job Description Extractor (LLM-driven)

• Parses the fetched content into structured JD fields:

• Requirements

• Responsibilities

• Skills

• Seniority

• etc.

---

3. Prepare and Aggregate

• Set Resume node

• Normalizes the resume into a clean JSON object.

• Merge/Aggregate node

• Builds a single payload containing:

{ "resume": "...", "job_description": "...", "meta": "..." }

---

4. AI Evaluation

• Recruiter Agent (LangChain node, powered by Google Gemini)

• Receives aggregated payload

• Returns a strict JSON-formatted screening report including:

• candidate_strengths

• candidate_weaknesses

• risk

• reward

• overall_fit_rating (0–10 numeric)

• justification

• Structured Output Parser

• Enforces JSON schema

• Ensures predictable downstream data

---

5. Identity Extraction & Logging

• Contact Info Extractor

• Extracts:

• Name

• Email

• Append to Google Sheets

• Writes:

• Date

• Name

• Email

• Strengths

• Weaknesses

• Risk

• Reward

• Justification

• Overall Fit Score

---

6. (Optional) Notifications / Follow-Ups

• Add Slack / Email / Webhook nodes
• Trigger alerts for high-fit candidates

---

Quick Setup Guide

👉 Demo & Setup Video 👉 Sheet Template 👉 Course

Nodes of Interest You Can Edit

Trigger — On Form Submission

• Change webhook URL
• Modify accepted form fields
• Add metadata capture (job_id, source)

Resume Extraction (Extract from File)

• Enable OCR fallback
• Adjust encoding/charset handling
• Replace with third-party resume parser

HTTP Request (Fetch Job Description)

• Configure timeouts
• Add retry policy
• Set headers
• Restrict allowed domains

Job Description Extractor (Information Extractor1)

• Modify extractor prompt/schema
• Add fields like must_have and nice_to_have

Set Resume (Prepare Resume)

• Strip headers/footers
• Normalize dates
• Split resume sections

Merge / Aggregate

• Modify payload structure
• Add context fields (job_id, recruiter_notes, source_platform)

Recruiter Agent (LangChain Agent)

• Edit system/user prompts
• Adjust model temperature
• Modify token limits
• Switch LLM provider

Structured Output Parser

• Update JSON schema
• Add fields like:
• experience_years
• certifications
• notice_period

Contact Info Extractor

• Add:
• Phone
• LinkedIn
• Location

Append to Google Sheets

• Modify column mapping
• Add fields like:
• workflow_run_id
• resume_link

---

What You’ll Need (Credentials)

• Google Sheets API credentials (OAuth or Service Account)
• Google Drive / Storage credentials (if resumes are stored there)
• LLM provider credentials (e.g., Google Gemini API key/service account)
• (Optional) OCR / Vision API credentials for scanned PDFs
• (Optional) Email / Slack / Teams webhook or SMTP credentials
• Access to public JD URLs (or credentials if behind authentication)

---

Recommended Settings & Best Practices

• LLM temperature: 0.0–0.3 for consistent output
• Max tokens: 800–1200 for justification (with enforced limits)
• Strict JSON schema: Fail fast on invalid structure
• Retries & timeouts:
• ~10s HTTP timeout
• 2 retries with exponential backoff
• Rate limiting: Protect LLM quotas
• Deduplication: Check existing email or resume hash
• Least privilege: Scope Google service account to target sheet only
• PII handling: Limit exposed fields; encrypt sensitive data if needed
• Schema versioning: Add schema_version column
• Error logging: Use Catch node with workflow_run_id
• Human review gate: Route borderline scores (6–7) for manual review

---

Customization Ideas

• Conditional alerts (overall_fit_rating >= 8)
• Multi-model scoring (Gemini + alternative model)
• Automated outreach emails
• ATS integration (Greenhouse, Lever, etc.)
• JD template library
• Multi-language resume routing
• Skill-level mapping (e.g., python: 4/5)
• Candidate scoring dashboard
• Resume storage with secure links

---

Troubleshooting — Quick Tips

Resume Extraction Issues

• Validate binary input
• Enable OCR for scanned PDFs
• Check encoding and file type

JD Fetch Failure

• Validate URL reachability
• Add headers (User-Agent)
• Increase timeout
• Provide auth if needed

LLM JSON Errors

• Lower temperature (0–0.2)
• Enforce strict JSON prompt
• Add retry with "fix-json" prompt
• Inspect raw LLM output

Google Sheets Append Fails

• Check credential expiry
• Confirm sheet ID and gid
• Validate column mapping
• Monitor API quota

Duplicate Rows

• Add email-based dedupe logic
• Hash resume content

PII Exposure

• Audit sheet sharing settings
• Use restricted service accounts

---

Tags / Suggested Listing Fields

• recruiting
• resume-parser
• ai-screening
• langchain
• google-gemini
• google-sheets
• n8n
• ats-integration
• pii-sensitive
• automation