n8n Headhunter Workflow – LLM Agent for Candidate Search

Overview

An intelligent recruitment automation system built on n8n that uses LLM agents to transform natural language hiring requirements into structured candidate searches, automatically enriching profiles from multiple sources and providing full transparency into search decisions.

The Problem

Traditional Recruitment Challenges

High Cost:

• LinkedIn Recruiter: $800+ per month per seat
• Additional sourcing tools: hundreds more
• Scale limitations based on budget
• Expensive for startups and small teams

Time-Intensive Process:

• Manual searches take hours
• Repetitive filtering and screening
• Multiple platform switching
• Difficult to maintain consistency

Lack of Transparency:

• Black-box search algorithms
• Can’t audit why candidates were selected
• Difficult to reproduce searches
• No insight into decision-making process

Data Quality Issues:

• Stale profile information
• Incomplete data across platforms
• Manual enrichment required
• Inconsistent data formats

The Solution

Intelligent Agent-Based Search

Natural Language to Structured Search:

Example Query:

"Find senior full-stack engineers in Singapore with 
React and Node.js experience, preferably from tech 
startups, who have worked on B2B SaaS products"

Agent Processing:

1. Parse: Extract requirements (skills, location, experience, industry)
2. Structure: Convert to database filters and search queries
3. Search: Query multiple data sources in parallel
4. Enrich: Gather additional profile information
5. Validate: Check candidates against criteria
6. Rank: Order by relevance
7. Return: Top 20 candidates with full audit trail

Time: Average 2 minutes from query to results

Multi-Source Search Strategy

Primary Source: People Dataset (200M+ entries)

Fast Filter-Based Retrieval:

• Structured data queries
• 5-10 seconds per batch
• High-quality baseline data
• Supports complex filtering

Supplementary: Web Search

Google Search (via Serper):

• Fresh, real-time results
• Finds profiles not in dataset
• LinkedIn dorking techniques
• Cost-effective coverage

Yahoo Search:

• Parallel search for broader coverage
• Alternative perspective
• Redundancy and validation
• Extended reach

Apify Enrichment:

• Profile detail extraction
• LinkedIn data scraping
• Contact information discovery
• $0.05 per profile cost

Combined Approach Benefits:

• Fast initial results from database
• Comprehensive coverage from web search
• Fresh data from recent updates
• Cost-optimized sourcing

LangChain-Stabilized Tool Calling

Problem with Standard LLM Agents:

• Random early termination
• Inconsistent tool usage
• Failure to complete search criteria
• Unpredictable behavior

LangChain Code Node Implementation:

Ensures Agent Loops Until:

• Target number of candidates found
• All search sources exhausted
• Explicit completion criteria met
• No valid candidates remain

Stability Improvements:

• 20% accuracy improvement
• Consistent search completion
• Reliable result counts
• Predictable execution

Technical Approach:

# Conceptual implementation
while len(candidates) < target_count and not exhausted:
    # Agent tool calling
    results = agent.run(search_query)
    
    # Validation
    validated = validate_candidates(results)
    
    # Accumulation
    candidates.extend(validated)
    
    # Progress tracking
    log_progress(candidates, target_count)

Full Auditability with Supabase Logging

Every Agent Action Logged:

Tracked Events:

• Search queries executed
• Data sources accessed
• Validation decisions
• Accept/reject reasons
• Enrichment steps
• Ranking criteria

Logged to Supabase:

• Real-time database writes
• Structured event data
• Timestamp tracking
• Session correlation

Webhook Exposure for Live Progress UI:

Solves “Why is Nothing Happening?” Problem:

Live Progress Dashboard Shows:

• Current search phase
• Candidates found so far
• Sources being queried
• Validation in progress
• Estimated time remaining

Benefits:

• User confidence during wait time
• Transparency into process
• Debugging capability
• Performance monitoring
• 90% pipeline auditability

Example Progress Updates:

[12:00:01] Parsing query...
[12:00:02] Searching database for 'React' + 'Singapore'...
[12:00:08] Found 45 candidates in database
[12:00:09] Launching Google search...
[12:00:15] Enriching top 30 profiles via Apify...
[12:00:45] Validating against criteria...
[12:01:23] Ranking by relevance...
[12:01:30] Top 20 candidates ready

Architecture

Workflow Components

n8n Orchestration:

Trigger (Natural Language Query)
    ↓
LLM Parser (Extract Requirements)
    ↓
Parallel Search Branches:
    ├─ Database Query (200M+ profiles)
    ├─ Google Search (via Serper)
    └─ Yahoo Search
    ↓
Result Aggregation
    ↓
Apify Enrichment ($0.05/profile)
    ↓
LangChain Agent Validation
    ↓
Supabase Logging + Webhook
    ↓
Ranking & Filtering
    ↓
Top 20 Candidates Output

Technology Stack

Automation Platform:

• n8n: Workflow orchestration
• LangChain Code Nodes: Agent logic
• Webhooks: API endpoints

Data Sources:

• People Dataset: 200M+ profile database
• Google Search: Serper API
• Yahoo Search: Direct API
• LinkedIn: Via Apify scraping

Storage & Logging:

• Supabase: Event logging and progress tracking
• PostgreSQL: Candidate storage
• Vector Database: (Planned) Semantic search

LLM & AI:

• OpenAI GPT: Query parsing and validation
• LangChain: Agent framework
• Embeddings: (Planned) Semantic matching

Search Pipeline

Phase 1: Query Understanding

• Natural language input
• LLM extracts structured requirements
• Validation of search criteria

Phase 2: Multi-Source Search

• Database: Fast structured queries
• Google: Fresh public profiles
• Yahoo: Extended coverage
• Parallel execution for speed

Phase 3: Profile Enrichment

• Apify LinkedIn scraping
• Contact information discovery
• Experience validation
• Skill verification

Phase 4: Validation & Filtering

• LangChain agent evaluates candidates
• Checks against original requirements
• Logs accept/reject decisions
• Maintains audit trail

Phase 5: Ranking & Delivery

• Relevance scoring
• Experience weighting
• Recency consideration
• Top 20 selection

Key Features

1. Natural Language Queries

• Conversational search input
• No need to learn query syntax
• Flexible requirement expression
• Context understanding

2. Multi-Source Intelligence

• 200M+ profile database
• Real-time web search
• LinkedIn enrichment
• Comprehensive coverage

3. Cost Optimization

• $0.05 per enriched profile
• Free database queries
• Efficient API usage
• Massive savings vs. LinkedIn Recruiter

4. Stable Agent Execution

• LangChain-powered reliability
• Guaranteed completion
• Consistent behavior
• 20% accuracy improvement

5. Full Auditability

• Complete action logging
• Decision transparency
• Reproducible searches
• Performance analytics

6. Live Progress Tracking

• Real-time status updates
• Webhook-powered UI
• User confidence
• 90% pipeline visibility

Performance Metrics

Speed:

• Average search time: 2 minutes
• Database queries: 5-10 seconds
• Enrichment: 30-60 seconds
• Total end-to-end: ~2 minutes

Cost:

• Profile enrichment: $0.05 each
• Monthly cost: Fraction of LinkedIn Recruiter
• Scalable pricing model
• Transparent cost tracking

Accuracy:

• 20% improvement with LangChain stabilization
• Consistent result quality
• Reliable candidate matching
• Reproducible searches

Coverage:

• 200M+ profiles in database
• Unlimited web search results
• Fresh data via Apify
• Comprehensive candidate pool

Transparency:

• 90% pipeline auditability
• Complete action logs
• Decision explanations
• Search reproducibility

Use Cases

1. Startup Recruiting

• Cost-effective sourcing
• Fast candidate discovery
• Flexible search criteria
• No expensive subscriptions

2. Agency Recruiting

• Batch candidate searches
• Multi-client support
• Transparent reporting
• Scalable operations

3. Internal HR

• Talent pipeline building
• Proactive sourcing
• Market research
• Competitive intelligence

4. Freelance Recruiters

• Low barrier to entry
• Pay-per-use model
• Professional results
• Client transparency

Future Development

Planned Enhancements:

Semantic Search & RAG

• Vector embeddings for profiles
• Natural language matching
• Beyond keyword search
• Contextual understanding

Stored Profile Intelligence

• Historical search results
• Profile versioning
• Trend analysis
• Relationship mapping

Advanced Filtering

• Culture fit scoring
• Salary expectation matching
• Availability detection
• Geographic optimization

Integration Expansion

• ATS integration
• Email outreach automation
• Interview scheduling
• CRM synchronization

Machine Learning

• Ranking optimization
• Success prediction
• Preference learning
• Automated improvement

Technical Insights

LangChain Agent Design

Tool Calling Stability:

• Explicit loop conditions
• Progress validation
• Error recovery
• State management

Decision Logging:

• Structured event format
• Timestamp precision
• Context preservation
• Audit trail maintenance

n8n Workflow Patterns

Parallel Execution:

• Simultaneous source queries
• Race condition handling
• Result aggregation
• Error tolerance

Webhook Architecture:

• Real-time updates
• Client subscriptions
• Event streaming
• Progress broadcasting

Data Pipeline Optimization

Batch Processing:

• Efficient API usage
• Rate limit management
• Cost optimization
• Performance tuning

Caching Strategy:

• Frequent query results
• Profile data persistence
• Search result storage
• Update scheduling

Impact

For Recruiters:

• Dramatically reduced costs (vs. $800+/month)
• Faster candidate sourcing (2 min vs. hours)
• Complete search transparency
• Reproducible results

For Candidates:

• Discoverable through multiple channels
• Fresh profile data considered
• Fair evaluation criteria
• Transparent selection process

For Organizations:

• Scalable recruitment operations
• Data-driven decision making
• Cost predictability
• Process standardization

Technologies Used

Automation:

• n8n workflow platform
• LangChain agent framework
• Webhook integrations

Data Sources:

• 200M+ profile database
• Google Search (Serper API)
• Yahoo Search
• Apify LinkedIn scraping

Storage:

• Supabase PostgreSQL
• Event logging system
• Real-time databases

AI/ML:

• OpenAI GPT models
• Natural language processing
• Future: Vector embeddings

Conclusion

The n8n Headhunter Workflow demonstrates how LLM agents, combined with multi-source data aggregation and complete auditability, can transform recruitment from an expensive, opaque manual process into a fast, transparent, and cost-effective automated system. The 90% auditability and live progress tracking set new standards for recruitment technology transparency.