[🔴 Closed] Round of 16 Challenge -Processing: Graph Data Modeling, Structure, and Importing

[Ari_Neo4j (Neo4j Community Manager)]

Round of 16 Challenge: "Sweet Sixteen" (July 9 to 15, 2026)

The bracket is heating up! This round we shift gears from Graph Data Science to the foundations that make every great graph possible: modeling, structure, and importing.

This Round's GraphAcademy Topic: Processing

This round is powered by the Processing category (Graph Data Modeling, Structure, and Importing), featuring 5 potential courses:

• Cypher Indexes and Constraints (2h): make your graph performant
• Importing CSV Data into Neo4j (1h): get data in with Cypher
• Introduction to Vector Indexes and Unstructured Data (1h): search the unstructured
• Building Knowledge Graphs with LLMs (1h): GenAI meets graphs
• Constructing Knowledge Graphs with Neo4j GraphRAG for Python (2h): Python + LLM pipelines

To qualify: complete at least 2 of the 5 courses from the Processing category (badges must be visible on your public GraphAcademy profile).

The Rule of 16: Maximums Edition

Last round, 32 was your target. This round, 16 is your ceiling. Constraints breed creativity, so your entire project must stay within ALL of these limits:

• Maximum 16 node labels and maximum 16 relationship types
• Maximum 16 properties on any single node or relationship
• Maximum 16 indexes + constraints combined
• Maximum 16 source files in your import (CSVs, documents for KG extraction, etc.)

Small model, big impact. The best entries will squeeze the most insight out of the tightest structure. Show us what elegant modeling looks like when you can't just add another label!

What to Build

Any project that showcases what you learned in the Processing courses: a well-modeled graph imported from real CSV data, a knowledge graph extracted from unstructured text with LLMs, a vector-search-powered app, or a combination. Real data is strongly encouraged. Show us your data model and how you got the data in!

Submission Template

Post a new topic titled Round of 16: [Project Name] in [Country] containing:

• GraphAcademy Cup Team Profile Link:
• GraphAcademy Public Profile Username: (anonymous profiles are not eligible for prizes!)
• Country:
• Courses Completed (minimum 2 from Processing):
• Project Name:
• Project Description (32 to 320 words):
• How your project stays within the Rule of 16 maximums:
• Screenshots (data model diagram encouraged!):
• GitHub / Demo URL:
• Dataset URL:
• Additional Notes:

Prize: Winner's Choice!

The Round of 16 winner picks their LEGO legend:

• Cristiano Ronaldo Display Figure OR

• Lionel Messi Display Figure

The eternal GOAT debate, settled brick by brick. Which icon will you build?

(Reminder: anonymous GraphAcademy profiles are not eligible for prizes!)

Deadline

Submissions close July 15, 2026, 23:59 (anywhere on Earth). Winners advance to the Quarterfinals!

Good luck, everyone! May your constraints never be violated and your imports never fail!

[prashant7090 (Prashant Sable)]

• GraphAcademy Cup Team Profile Link: India | GraphAcademy

• GraphAcademy Public Profile Username: Prashant Sable | GraphAcademy

• Country: India

• Courses Completed Yes , 3

  

• Project Name: BBC NewsGraph Semantic Intelligence

• Project Description : BBC NewsGraph Semantic Intelligence is an AI-powered semantic search system that combines Neo4j graph database and local embeddings to discover BBC News articles based on meaning rather than keywords. It uses vector search to retrieve contextually similar articles and provides a foundation for future GraphRAG applications.

• How your project stays within the Rule of 16 maximums: attached screenshots in github repos, Total node labels, relationships, constraints and index are under 16

  Node Label	Purpose
  Article	Stores BBC news articles and embeddings
  Category	Stores news categories

  ---

  Relationship Types

  Relationship	Direction	Purpose
  BELONGS_TO	(Article)-[:BELONGS_TO]->(Category)	Connects articles to their news category

  ---

  Constraints

  

  Index:

  

• Screenshots : data model -

  

• GitHub / Demo URL: GitHub - prashant7090/bbc-news · GitHub

• Dataset URL: BBC News Archive | Kaggle - Real Data:

• **Additional Notes:

  What I Learned:**

• Imported and structured BBC News data into Neo4j using a graph-based data model.

• Designed nodes, relationships, constraints, and indexes to maintain data consistency and improve query performance.

• Generated local embeddings using all-MiniLM-L6-v2 and stored vector representations with Article nodes.

• Implemented Neo4j Vector Search to perform semantic search based on meaning rather than keyword matching.

• Learned the foundation of building AI-powered search systems using Graph Database + Embeddings.**

  Embeddings & Semantic Search:**
  Article content is converted into 384-dimensional vector embeddings using the all-MiniLM-L6-v2 model and stored in Neo4j. Semantic search converts user queries into vectors and uses Neo4j Vector Index with cosine similarity to retrieve contextually similar articles beyond keyword match

  Create local embedding:

  

  Verify embedding:

  

  Semantic search:

  Question (Business Category): "Companies reporting financial growth"

  

  **
  Question (Cross Category): "Discover articles where innovation created new revenue opportunities"**

  

  **
  Question (Politics Category): Find stories about international political relationships.**

  

  **
  Question(Entertainments category): Find news about successful films and their impact
  **

  

Question (Health Category): find articles about discoveries improving people's lives.

[shubhamsingh.41188]

Round of 16: [CodeMind Graph AI — Intelligent Codebase Knowledge Graph] in [India] containing:

• GraphAcademy Cup Team Profile Link:India | GraphAcademy Cup

• GraphAcademy Public Profile Username: shubhamsingh.41188

• Country: India

• Courses Completed (minimum 2 from Processing):

  Cypher Indexes and Constraints (2h) Building Knowledge Graphs with LLMs (1h) Constructing Knowledge Graphs with Neo4j GraphRAG for Python (2h)

• Project Name:CodeMind Graph AI — Intelligent Codebase Knowledge Graph

Project Description (32 to 320 words):

1. CodeMind Graph AI transforms codebases into intelligent Neo4j knowledge graphs.
2. It parses files, builds nodes (Files, Functions, Imports, Projects), and creates relationships
3. (CONTAINS, IMPORTS, CALLS). Vector embeddings (768-dim) from Ollama enable semantic search.
4. GraphRAG with llama3.1:8b answers natural language queries like "Find functions related to chatbot."
5. 6 GDS algorithms (PageRank, Degree, Betweenness, Louvain, Node Similarity, WCC) reveal hidden
6. patterns. Built within Rule of 16: exactly 15 indexes+constraints (7 uniqueness + 7 range + 1 vector).

How your project stays within the Rule of 16 maximums:

• Node Labels ≤16 5 (Project, File, Folder, Function, Import)
• Relationship Types ≤16 3 (CONTAINS, IMPORTS, CALLS)
• Max Properties per Node ≤16 6 (name, path, language, size, docstring, embedding)
• Max Properties per Relationship ≤16 0 (no properties on relationships)
• Indexes + Constraints Combined ≤16 15 (7 constraints + 7 range indexes + 1 vector index)
• Source Files ≤16 7 files imported from GitHub​

Screenshots (data model diagram encouraged!)

1. GitHub Repository Load/Local File Path

User Action: Entered GitHub URL And Local File Path → Clicked "Parse & Load"

Shows: System fetched code, parsed 7 files, found 5 functions, 25 imports, 64 relationships → Graph built!

Caption: "Loading codebase from GitHub — parsing and building initial graph"

2. Graph Build + Embeddings

User Action: After parsing, system automatically generated embeddings

Shows: 5 code elements embedded → 39 total nodes created
Caption: "Graph created with 39 nodes — embeddings generated for semantic search"

3.Node Labels Distribution

Shows: Import (25), File (7), Function (5), Project (1), Folder (1) = 39 nodes

Caption: "5 node labels — all under Rule of 16 limit"

4. Relationship Types Distribution

Shows: IMPORTS (25), CONTAINS (13), CALLS (1) = 39 relationships

Caption: "3 relationship types — all under Rule of 16

5.Constraints + Indexes (7+7)

6Vector Index (1)

Shows: Neo4j vector index code_embedding — type: VECTOR, status: ONLINE

Caption: "Vector index created — enabling semantic search"

7**.Embedding**

Shows: A Function node has embedding property with 768-dimension array

Caption: "Vector embeddings stored — 768-dimension for each function"

8. Knowledge Graph

Graph Details:
- Total Nodes: 39
- Total Relationships: 39
- Node Labels: 5 (Project, File, Folder, Function, Import)
- Relationship Types: 3 (CONTAINS, IMPORTS, CALLS)
- Max Properties per Node: 6 (name, path, language, size, docstring, embedding)

What This Graph Shows:
- Files and their locations (File nodes)
- Functions defined in each file (Function nodes)
- Import statements (Import nodes)
- Project structure (Project, Folder nodes)

Vector Embeddings:

• 768-dim embeddings on Function nodes
• Code embedding vector index for semantic search

9. GDS Analytics Dashboard

Description:

GDS Analytics dashboard showing basic stats: 39 nodes, 39 relationships, 5 functions, 25 imports, 7 files, 0 classes, 0 methods. Shows GDS Analytics integrated.

10. Knowledge Graph + LLM

1.User asks: What files are in this project?

->Shows file → function relationship

2.User asks: "Who calls the function you just mentioned?"

->GraphRAG + Graph Traversal.

3.User asks: "Find functions related to 'chatbot'."

→Shows semantic search

GitHub / Demo URL:shubham123sing/CodeMind-Graph-AI

Dataset URL: streamlit/llm-examples: Streamlit LLM app examples for getting started

Additional Notes:

1. Key Features: Codebase Import, Knowledge Graph, Vector Embeddings, GraphRAG Chat,GDS Analytics (6 algorithms), Random Discovery, Memory Management, Fully Offline.
2. Future Scope: MCP Integration, Multi-Language Support, VS Code Extension
3. Coding AI Agent Ready:Graph provides persistent code memory — AI agents can query call chains, dependencies, and code structure instantly.