SVC—008 ● IN-PROG GRAPH AI EARLY 2026

Neo4j
Graph
Agent

Relational AI Traversal

A graph-based AI agent that maps complex entity relationships between jobs, skills, and candidates in Neo4j. Custom node architecture enables multi-hop Cypher traversal — allowing the agent to reason about skill gaps, career path adjacency, and candidate-to-role fit across thousands of interconnected entities.

Graph
DATABASE
Entity
MAPPING
Multi-hop
TRAVERSAL
Neo4jCypherGraph AIPythonLLM Orchestration
// UI PREVIEW — GRAPH INTELLIGENCE CONSOLE
NEO4J GRAPH AI AGENT Graph Console
In development
ENTITY EXTRACTION
JOBML Engineer248
JOBData Scientist193
SKILLPython1.2K
SKILLPyTorch487
SKILLSQL891
PERSONCandidate nodes3.4K
6.2K nodes · 18.7K edges
GRAPH TRAVERSAL
ML Eng Python PyTorch FastAPI SQL Spark AWS REQUIRES ADJACENT
SKILL GAP ANALYSIS
TRAVERSAL PATH
python →[ADJACENT]→ sql →[REQUIRED_BY]→ ml-engineer
SKILL GAP — CANDIDATE C-2041
PyTorch — missing
SQL — partial (2yr)
Python — strong (5yr)
2 gaps · 1 path to role
02 // ARCHITECTURE

The Pipeline

Job & Skill
Data
→
Entity
Extractor
→
Neo4j
Graph DB
→
Cypher
Queries
→
LLM Path
Reasoning
→
Skill Gap
Analysis
STEP 01
Entity Extraction
Job descriptions and candidate profiles parsed. Entities extracted: job titles, required skills, experience levels, and candidate attributes.
STEP 02
Node Creation
Each entity type becomes a Neo4j node. Custom schema: Job, Skill, Candidate, Domain. Relationships: REQUIRES, HAS_SKILL, ADJACENT_TO, LEADS_TO.
STEP 03
Graph Population
Real-world job and skill data ingested into Neo4j. Edge weights represent frequency and strength of relationships across the dataset.
STEP 04
Cypher Traversal
LLM generates Cypher queries from natural language. Multi-hop traversal finds shortest paths, identifies skill clusters, and surfaces role adjacencies.
STEP 05
LLM Reasoning
Traversal results fed to LLM for interpretation. Agent explains skill gaps, suggests learning paths, and ranks candidate-to-role fit from graph evidence.
03 // STACK

Built with

GRAPH DATABASE
Neo4j
Native graph storage with ACID transactions. Optimized for relationship traversal — queries that would require multiple JOINs in SQL run in a single Cypher hop.
QUERY LANGUAGE
Cypher
Declarative graph query language. LLM generates Cypher for complex multi-hop patterns: MATCH (job)-[:REQUIRES]->(skill)-[:ADJACENT]->(related_skill).
ENTITY MAPPING
Graph AI
Custom node schema designed for the HR domain. Bidirectional edges encode both directions of skill-job relationships for flexible traversal.
REASONING
LLM Orchestration
LLM layer translates natural language queries into Cypher, interprets traversal results, and generates human-readable skill gap reports.
ANALYSIS
Skill Gap Engine
Compares candidate skill subgraph against job requirement subgraph. Scores gap severity and recommends adjacent skills as bridging steps.
BACKEND
Python
Neo4j Python driver for query execution. LangChain agent wraps the graph tool for conversational querying over the entity relationship graph.
See the rest
of the deployments.
← ALL PROJECTS