Ship.AI App

Inspiration

Shipping powers global trade, but most maritime teams rely on outdated tools, scattered data, and manual workflows. We built Ship.AI App to bring intelligent, real-time awareness to the sea - using AI agents to make ports smarter, faster, and safer.

What it does

• Tracks ships in real time on a 3D globe
  
• Click any ship to get instant details
  
• AI chat understands ship context and answers questions
  

How we built it

• Frontend: React, TypeScript, Tailwind (glassmorphic UI)
  
• Maps: Mapbox
  
• AI: Claude 4 Sonnet via AWS Bedrock
  
• Data: SQLite local DB + MCP client
  

Challenges

• Maintaining consistent AI context across dynamic user interactions
  
• Designing real-time triggers with low latency and no false positives
  
• Parsing and reasoning over unstructured maritime documents
  
• Ensuring smooth 3D rendering and performance with live ship data

Accomplishments

• Built a real-time AI agent for maritime tracking and analysis
  
• Integrated voice, document parsing, and automation into one workflow
  

What we learned

• How to design vertical AI agents using tool chaining
  
• Claude 4 effectively handles contextual maritime data
  
• Voice interfaces enhance usability in operational settings

What's next

• Expand to historical tracking and predictive modeling
  
• Improve maritime-specific AI responses through fine-tuning
  
• Extend the platform to support other verticals such as logistics and aviation
• Voice interaction through MiniMax Audio for hands-free operation
  
• Real-time event triggers using Inngest AgentKit (e.g. ship inactivity alerts)
  
• Response validation and logging via Operant AI
  

Built With

• claude-4-via-aws-bedrock
• glassmorphic
• mapbox
• operant-ai-(gatekeeper-&-woodpecker)
• react
• sqlite
• tailwindcss
• typescript