Onboard Local LLM · IHO S-100 · Voice-Commanded

enCaptain AI

The Onboard Second Officer for Safer Voyages

An AI-powered, voice-commanded navigation copilot that lets an onboard AI read, query, and reason over next-generation IHO S-100 electronic charts — fully onboard, cloud-free.

Watch Demo Demo Video

ZERO EXTERNAL CONNECTIVITY 8 S-100 PRODUCT SPECS 10,000 TARGETS / 30 ms

A solo skipper on a small fishing boat using enCaptain AI on a tablet

The Problem

Where AI is needed most, AI is absent.

Risk at sea concentrates where officers are overloaded, alone, or far from help — precisely where cloud-based assistants cannot follow.

~75%

Human error drives maritime accidents

Industry analyses attribute roughly 75% of maritime accidents to human error (Allianz Safety & Shipping Review) — and casualties ripple through coastal supply chains.

0 bars

Cloud AI cannot follow

Cloud-based assistants are unreliable mid-ocean and cannot operate inside security-isolated naval networks — the environments that need a second officer most.

2029

The S-100 transition is coming

The IHO S-57-to-S-100 chart transition, required for new ECDIS installations from 2029, forces costly fleet upgrade decisions. Today's ECDIS offers rules and alarms — not AI chart interaction.

Concept Flow

Your chart is no longer a picture.

S-101 chart context
MCP context — 13 purpose-built query tools
enCaptain AI — locally hosted LLM reasoning
Operational action — route, alert, voice response

One watch, full situational awareness.

The world's first S-101-to-MCP pipeline turns the electronic chart into a queryable knowledge base that an onboard AI can reason over through natural language.

enCaptain AI concept flow — S-101 chart context, MCP context, enCaptain AI, operational action

Capabilities

Command, alert and response — in one view.

A tireless AI second officer for every vessel, from the largest merchant ship to the smallest fishing boat.

01 / Ask the chart

Natural-language chart queries, offline

Interactive, spoken queries against onboard S-100 data — answered by voice in seconds, with zero external connectivity.

"Any obstruction shallower than my draft within two miles?"

02 / Vessel-specific routes

Physics-grounded route planning

The hybrid planner (D* Lite + ONNX TabNet, trained on 1,082 real ferry routes) minimizes physical risk from your vessel's own draft, hull dynamics, and forecast sea state — not just distance.

03 / One operational view

Eight S-100 layers, one tactical display

Concurrent rendering of S-101/102/104/111/121/122/124/413 alongside legacy S-57/S-52, following IHO S-52 presentation rules and MIL-STD-2525C/D symbology — readable with the training officers already have.

04 / Keep existing equipment

Software-only S-100 support

Delivered as software with C++, C#, Java, and React.js APIs, it adds seamless S-100 support to compatible legacy bridges — no hardware replacement, less electronic waste.

In Operation

From route selection to safe execution.

enCaptain AI voice dialogue during navigation with hazard alarms and weather briefing — **Voice command workflow**Ask, hear, acknowledge — route planning and hazard queries handled end-to-end without a single keystroke.

Route visualization with vessel tracking and hazard-aware navigation on S-100 environment layers — **Hazard-aware navigation**Route visualization, vessel tracking, and live operational context on fused S-100 environment layers.

Ship dynamics panel with CFD hull pressure, under-keel clearance and XTE alarms — **Physics-aware insight**6-DOF hull motion, CFD hull pressure, under-keel clearance, and cross-track alarms — computed every simulation step.

Environment data layers — wind, current, wave, storm surge — fused on the electronic chart — **Live environment fusion**CMEMS and NOAA GFS wind, current, wave, and surge layers refreshed six-hourly — or pre-loaded for air-gapped voyages.

Engineering

A full LLM agent inside an air-gapped vessel.

Where cloud dependence kept generative AI out of security-isolated maritime environments, enCaptain AI runs its entire generative-AI stack onboard.

S-101 → MCPChart as an AI tool

S-101 datasets (ISO 8211) are parsed into feature objects, converted to GeoJSON, embedded as 384-dimensional vectors, and indexed in FAISS/Milvus. An MCP server exposes 13 query tools — hazard search, depth statistics, location-based feature extraction — to the local LLM.

Voice → ActionOn-device speech loop

On-device Whisper converts voice to intent; the agent automatically calls chart and weather tools and replies via Transformer TTS (WebAssembly). The LLM is served by vLLM on an onboard edge server; tablets and bridge consoles connect as clients.

Local StackZero external connectivity

Connected mode refreshes weather (CMEMS, NOAA GFS) six-hourly; air-gapped mode uses forecasts pre-loaded before departure. All inference, vector search, and rendering run with zero external connectivity — voice data never leaves the ship.

Safety architecture. Every safety-critical route and hazard output is verified against charted depths and hazards by a deterministic rule layer before display. Safety queries use exact geospatial computation, not semantic search — and the officer retains final authority, always.

Proven & Certified

Built on a chart engine already at sea.

200+

chart engine units delivered to defense integrators & active naval platforms

Korean patents mapped to product functions

GS 16-0363

software quality certification (TTA)

registered intellectual property rights

ENDCS's underlying electronic chart engine already powers HD Hyundai Avikus' HiNAS and Samsung Heavy Industries' SAS. Key patents include No. 10-2722407 (S-100 hydrographic modeling powering the chart-as-tool pipeline), No. 10-1982084 (deep-learning route generation), and No. 10-2584065 / 10-2556481 (real-time vessel-sinking disaster response underlying hazard alerting).

Technical Specifications

enCaptain AI v1.0

Rendering: 10,000 dynamic targets within 30 ms; eight S-100 product specifications (S-101/102/104/111/121/122/124/413) plus legacy S-57/S-52; MIL-STD-2525C/D symbology
AI Stack: vLLM-served local LLM · Whisper STT · Transformer TTS (WebAssembly, on-device) · FAISS/Milvus vector search · 13-tool MCP server
Route Planning: D* Lite pathfinding + ONNX TabNet risk model (trained on 1,082 real ferry routes); risk thresholds configurable per vessel type, draft, and speed
Hull & Physics: 6-DOF hull-motion models; CFD (OpenFOAM); under-keel clearance monitoring with configurable margin (default 2.0 m)
Weather: ECMWF, NOAA GFS, CMEMS (6-hour refresh); pre-loaded forecasts for air-gapped operation
Deployment: Air-gapped Kubernetes (Rancher/Helm) on RedHat Linux or bare metal; from a single tablet client to shipboard clusters
Platforms & APIs: Windows, Linux, iOS, Android · APIs for C++, C#, Java, React.js · NMEA 0183, AIS, MarineTraffic integration
Distribution: B2B/B2G — OEM SDK/API licensing to ECDIS manufacturers and shipbuilders; subscription from USD 10,000 per vessel per year (enterprise); small-vessel tier available

Human Security for All

Safe navigation at sea should not be a privilege of the largest ships. enCaptain AI puts a second officer on every vessel — down to the fishing boats that need one most.

enCaptain AI was born from tragedy: a small-vessel accident on the Korean coast where no one was watching the chart. Our patented real-time vessel-sinking disaster response technology began as an answer to that loss. All AI runs onboard — voice data and route queries never leave the vessel: privacy by architecture, not by policy. Voice interaction makes safe navigation accessible to aging skippers and crews with limited literacy. Final authority always rests with the mariner, not the AI.

Get in Touch

Put a tireless second officer on your bridge.

enCaptain AI is deployed B2B/B2G through shipbuilders, shipping companies, navies, and ECDIS manufacturers. Contact us for a live demonstration, SDK/API licensing, or fleet evaluation.

Watch Demo Demo Video

Company: ENDCS Co., Ltd.
Address: 513HO, 10, Yatap-ro 81beon-gil, Bundang-gu, Seongnam-si, Gyeonggi-do 13497, Republic of Korea
Email: bhhwang@endcs.net
Web: www.endcs.net
Product: naru.endcs.net