Sampled Graph Visualization
Gradio Output
Sample Graph Visualisation in ArangoDB UI

About the Project: ArangoLens AI 🌟

🚔 Inspiration

The project was inspired by the growing need for intelligent crime analysis systems in metropolitan areas. With crime data being generated at massive scales, law enforcement agencies require advanced systems that can not only store and retrieve information but also reason and generate meaningful insights.

The Chicago Crime Records Dataset provided a rich, real-world dataset to explore how GraphRAG could revolutionize crime data analytics. Our objective was to leverage:
✅ GraphRAG (Graph-based Retrieval-Augmented Generation)
✅ GPU-accelerated graph analytics

By combining these, we built an Agentic Application capable of answering natural language queries related to crimes, criminals, locations, and other connected information with high precision.

🔎 What It Does

ArangoLensAI is an AI-powered crime intelligence agent that allows users to query, analyze, and visualize crime patterns using natural language. It integrates GraphRAG, ArangoDB, and GPU-accelerated graph analytics to generate structured and interpretable insights from complex crime networks.

🏙️ With ArangoLensAI, users can:

📍 Find crime hotspots based on location and time trends.
🔗 Identify criminal associations through graph-based relationship analysis.
📅 Retrieve historical crime records and related insights using natural language queries.
🛣️ Run shortest path analysis between criminals, crime types, and districts.

This solution enhances law enforcement investigations by providing an interactive and AI-driven approach to crime intelligence.

🏗️ How We Built It

To transform raw crime data into an intelligent AI-powered crime analysis system, we followed a structured approach integrating graph databases, GPU acceleration, and natural language processing.

🗂️ 1. Data Preparation

📥 Extracted crime data from the Chicago Crime Records Dataset.
🛠️ Cleaned and filtered the dataset to remove inconsistencies.
🎭 Added synthetic names to simulate real-world criminal tracking.

🔗 2. Graph Construction & Storage

📊 Converted crime data into a graph representation using NetworkX.
🗄️ Persisted the graph into ArangoDB, ensuring efficient query execution.
📈 Validated graph structure through centrality and connectivity analysis.

🤖 3. Agentic App Development

⚡ Built an LLM-powered agent using LangGraph.
🔍 Integrated AQL queries for structured data retrieval.
🚀 Leveraged cuGraph for fast, GPU-accelerated graph computations.
🖥️ Developed an interactive UI with Gradio for seamless user interaction.

⚠️ Challenges We Ran Into

🧹 Data Cleaning – The original dataset contained inconsistent and missing values.
🎮 GPU Integration – Setting up cuGraph with NetworkX for accelerated graph analytics.
🔄 Hybrid Query Optimization – Merging AQL queries with NetworkX analytics without compromising performance.
🚫 Hallucination Mitigation – Ensuring that the agent’s responses were factually correct and grounded in the graph data.

🏆 Accomplishments That We're Proud Of

✅ Successfully implemented GraphRAG, reducing AI hallucinations in retrieval tasks.
⚡ Optimized complex graph queries using a hybrid AQL + cuGraph approach.
📡 Built a scalable, multi-modal crime intelligence system that can be extended to real-world law enforcement.
⏳ Enabled real-time querying of crime relationships using natural language.
📊 Created a visually interactive experience that makes graph-based AI insights accessible to all users.

📚 What We Learned

Throughout the journey, we gained profound insights into:

🔗 Graph Data Modeling – Understanding how to represent crime data as nodes and edges for enhanced retrieval.
🤖 GraphRAG – Implementing graph-based RAG systems to mitigate hallucinations in AI responses.
📊 NetworkX & ArangoDB – Transforming tabular data into graph structures and persisting them into a graph database.
⚡ GPU-Accelerated Graph Analytics – Utilizing NVIDIA cuGraph for large-scale graph processing.
🏗️ Agentic Applications – Building interactive AI agents using LangGraph and AWS Bedrock for natural language querying.

🔮 What's Next for ArangoLensAI?

📊 Advanced Path Analysis – Identifying criminal networks and co-occurrence patterns.
📊 Generating Interactive Graphs – Based on results generating user interactive graphs
⏳ Real-Time Querying – Integrating streaming data pipelines for live crime analysis.