Skip to main content
Back to case studies
SeaSure logoClient projectLive

Marine · AI Agents

SeaSure Marine Maintenance Platform

Built by Van Data Team for SeaSure

Subscription platform pairing AI troubleshooting with a document vault and a vetted service-provider directory, with real-time chat and video support.

domain modules in production
8
subscription tiers with usage quotas
3
seasure.ai
SeaSure home page
SeaSure on a phone

Measured outcomes

  1. Outcome 01

    A specialised LangGraph agent per boat subsystem

  2. Outcome 02

    Three subscription tiers with usage quotas and rotating OAuth sessions

  3. Outcome 03

    Real-time chat and live video over a pgvector-backed service directory

Challenge pressure

Boat owners let scheduled maintenance slip because troubleshooting knowledge and service history lived in different places.

The situation

A vessel is several systems that fail in different ways, and its upkeep record is usually a folder of receipts. Owners postpone service because the diagnosis, the history, and the person who can carry out the work are never in the same place.

What we built

A FastAPI backend of eight domain modules with a specialised LangGraph agent per boat subsystem, routing across Claude 3 and GPT-4. The React 19 and TypeScript frontend runs roughly 25 pages on TanStack Query. PostgreSQL with pgvector carries 25 or more indexes with tuned connection pooling behind Alembic migrations. Stripe covers three tiers with usage quotas, authentication is JWT with Google OAuth and refresh-token rotation, and WebSocket chat and live video endpoints run on Google Cloud Run with SendGrid and Twilio alerting.

Constraints and decisions

One agent per subsystem rather than one agent for the whole boat: a general troubleshooter degrades as its surface widens, and per-subsystem routing keeps each agent narrow enough to stay accurate. Refresh-token rotation was required because the product is used from a moving vessel, where connections drop and resume constantly.

Technology stack

FastAPILangGraphReact 19PostgreSQLpgvectorStripeCloud RunTwilio

Share this

Need a similar system?

If this case study maps to your current workflow, the next useful step is a scoped project conversation with constraints, timeline, and integration needs.

Related case studies