CareerGraph

🥉 3rd place — Hack Night London by The Builders Collective at Tessl - May 13, 2026 Solo build - one evening - 39 teams - 150 people

CareerGraph is a graph-first AI career mentor. It turns a learner's career background into a persistent knowledge graph in Neo4j, then answers graph-native questions: which skill is blocking your progress, which project covers the most gaps, and what is your shortest path to your target role.

Unlike generic "learn X next" advice, every recommendation is justified by a graph traversal, and not a language model guess.

Demo

→ Open interactive career map (HTML)

Screenshots

HTML Career Map — color-coded graph of skills, evidence, projects, and career path:

Neo4j Browser — 22 nodes, 45 relationships live in the graph database:

CLI Demo — full pipeline output in the terminal:

Quick Demo

.venv/bin/python careergraph.py demo --demo-fast

The default path is deterministic and does not require API keys or Neo4j.

Local Installation

python3 -m venv .venv
.venv/bin/python -m pip install -r requirements.txt

The current local .env contains:

NEO4J_URI=bolt://localhost:7687
NEO4J_USERNAME=neo4j
NEO4J_PASSWORD=CareerGraphPass123
NEO4J_DATABASE=neo4j

OPENAI_API_KEY is intentionally blank. Add your key only when you want live LLM extraction.

Optional LLM Extraction

Set OPENAI_API_KEY in .env, then run:

.venv/bin/python careergraph.py demo --use-llm

If the LLM output fails validation, CareerGraph retries once and then falls back to the built-in Full-stack Developer graph.

Optional Neo4j Persistence

Start Neo4j with Docker:

scripts/start_neo4j.sh

Neo4j Browser opens at http://localhost:7474 — username neo4j, password CareerGraphPass123.

Write and query the graph:

.venv/bin/python careergraph.py demo --demo-fast --write-neo4j --query-neo4j

Stop Neo4j:

scripts/stop_neo4j.sh

Tests

.venv/bin/python -m unittest discover -s tests

Vision

CareerGraph was built in one evening as a working MVP — not a mock, not a slide deck. The goal was a system that actually runs, writes to Neo4j, and produces a real diagnosis end-to-end. The core architecture — graph-native reasoning over LLM-extracted knowledge — is designed to scale into a much richer career intelligence layer.

Richer Evidence Ingestion

• Resume and CV parsing to extract past roles, projects, and technologies as graph nodes
• GitHub activity analysis to infer proven skills from real commit and project history
• Job description parsing to auto-generate target career requirement graphs
• Course and certification records linked directly to skill nodes

Smarter Graph Reasoning

• Skill confidence scoring weighted by evidence recency and depth
• Timeline planning with estimated months-per-skill based on difficulty and prerequisites
• Multi-career comparison showing shortest-path divergence across N target roles
• Skill demand overlay connecting the graph to live job posting frequency data

Collaborative Knowledge

• Mentor annotation of learner graphs (mentors add evidence and prerequisite edges directly)
• Community-validated prerequisite chains (crowd-sourced skill ordering)
• Cohort comparison showing where peers sit in the same career graph

Product Surface

• Web UI over the same Neo4j layer with interactive graph exploration
• Natural language query interface ("what is blocking me from becoming a data engineer?")
• Export to personal learning roadmaps (Notion, Obsidian, markdown)
• API integration with learning platforms (Coursera, LinkedIn Learning, Udemy)