PROJECT HARRISON

We seek to make maritime knowledge more open, accessible, and useful by building artificial intelligence tools that solve persistent problems within the maritime industry. Our long-term vision is rooted in responsible innovation, open collaboration, and a deep commitment to maritime-related data science communities.

Our Values Project Harrison is committed to open, responsible, and ethical maritime innovation. We partner exclusively with organizations that share our values. We do not enter into partnerships, data agreements, or licensing arrangements with organizations whose activities involve maritime warfare, weapons systems, military targeting, surveillance for political repression, or any systematic restriction of human rights as defined under the United Nations Universal Declaration of Human Rights.

Agentic Application

MIRA — Maritime Informatics Reasoning Agent

MIRA is an interactive maritime application that lets users query voyages using geographic information systems and vetted government and NGO data.

  • Natural-language voyage queries
  • Interactive maps
  • External REST API for maritime research
  • Synthetic Statement of Facts™ (SSOF)
  • Utilizes a unique Voyage Context Protocol™ (VCP)
Pilot Program

Access MIRA

MIRA is currently available to approved pilot participants. Log in with your authorized account to query voyages, generate routing plans, and explore the full capabilities of the platform.

Log in to MIRA
Access is limited to pilot participants. Early access will be available late 2026.
-->
GitHub: searoute_mcp
Open Source

The First Open-source Model Context Protocol Server in the Commercial Maritime Industry

searoute_mcp is a Model Context Protocol (MCP) server for maritime routing. It integrates searoute-py with MCP, exposing tools that allow LLM clients to:

  • Compute oceangoing route distances in nautical miles
  • Retrieve full oceangoing routes with waypoints (GeoJSON)
  • Compare against great-circle (geodesic) distances

Explore the project, installation instructions, and usage examples on GitHub:

View on GitHub
Domain Protocol

Voyage Context Protocol (VCP)™

VCP is a domain protocol and maritime cognitive architecture for managing maritime GIS within an AI workflow. VCP brings route geometry, geospatial overlays, and nautical context into a structured voyage representation that a language model can interpret and use to solve operational problems with spatial and temporal fidelity.

  • Maritime GIS as first-class context: routes, waypoints, overlays, constraints, and spatial signals
  • Nautical science alignment: context shaped around real navigation and voyage operations
  • AI-ready structure: consistent representations that enable reliable reasoning and tool use
  • Composable problem-solving: supports iterative updates without losing the voyage narrative
  • Traceable updates: changes can be tracked and reviewed as the voyage context evolves
Why it matters
Most AI systems treat maps as pictures and voyages as text. VCP treats maritime GIS as structured context — so Generative AI can reason over routes, areas, and operational constraints in a way that's usable for decision support.
Notice: VCP™ is proprietary. Open-source components (e.g., GitHub repositories) are licensed under their respective repo licenses. Copyright © 2025 Project Harrison. All rights reserved.
Voyage Context Protocol (VCP)™
Operational maritime context for AI reasoning.
Publications

Research & White Papers

Pilot Evaluation PH-2026-001  ·  March 15, 2026
Evaluating Generative Artificial Intelligence: Maritime Route Intersections and Estimated Time of Arrival

We ask a basic question of generative artificial intelligence: what is the point and time of arrival of a line intersecting a maritime route? We measure outcome across accuracy, API cost, and speed — comparing two general-purpose models against MIRA, our agentic maritime reasoning system, across 50 real-world voyage plans.

MIRA  50 / 50  ·  $0.011 Gemini  13 / 50  ·  $0.079 Claude  44 / 50  ·  $1.596
Author: Jordan Taylor · Editor: Dr. Padmapriya Jayaraman
Read White Paper
Key Finding

General-purpose language models operating without domain-specific tooling may not be reliable instruments for maritime geospatial reasoning. Agentic systems with deterministic geometric tooling outperform them dramatically on domain-specific tasks.

Our Team

Leadership

Jordan Taylor
Jordan Taylor
Director

U.S. Coast Guard–licensed Chief Mate Unlimited and Fulbright Specialist (U.S. Department of State), recognized for peer-reviewed work in maritime software engineering and artificial intelligence. 25+ years in commercial marine transportation at sea and ashore, including 10+ years at sea on VLCCs. Former U.S. Navy Strategic Sealift Officer (SSO) and U.S. Navy Rescue Swimmer.

Dr. Padmapriya Jayaraman
Dr. Padmapriya Jayaraman
Senior Advisor

AI Scientist and Research Advisor specializing in computer vision, machine learning/deep learning, and agentic workflows. 8+ years across research and academia, leading projects from method development through deployment in maritime, agriculture, desalination/clean water, and sustainability. Principal Investigator on a Government of India–funded AI-based weed eradication initiative, with published research and patents in applied AI and autonomous systems.

Quinn Hoffman
Quinn Hoffman
Volunteer Software Engineer

Quinn Hoffman is a Software Engineer and Computer Programmer living and working in Portland, Oregon. After studying Computer Science at Oregon State University with a specialization in Simulation and Graphics programming, he is using his family ties to shipping to help create platforms to help maritime professionals accomplish their objectives.

Organization

Project Harrison is a California Non-Profit Public Benefit Corporation organized exclusively for educational and scientific purposes. We operate without profit motive — reinvesting all resources into open maritime and data science initiatives. We are independent in governance and accountable to the communities we serve.