Ship bridge at sunset with navigation instruments
Master Mariner · MSc · LLM

Maritime Engineering
& Quantitative Research

Production tools for weather routing, tanker cargo operations, and systematic equity research — built by a serving master mariner with 27 years at sea, 90 FPSO offloadings, and 750+ voyages operated from the shore side.

Open WindMar Publications
27 Years at Sea
4 Production Platforms
119+ Research Trials
22 Publications
Credit photo: SL Mar
Expertise

Four domains, one operational thread

Each surface is built from direct operational experience — ship performance physics, cargo hydraulics, onboard AI, and systematic research.

WindMar — typhoon avoidance routing with wind and wave height overlay

Weather Routing & Ship Performance

Physics-based ship modelling, A* pathfinding with safety constraints, and Monte Carlo fuel estimation across ECMWF/GFS ensembles. Vessel models calibrated against measured consumption at sea.

WindMar →
LCSMar cargo synoptic during a live loading — tanks filling, flow paths lit

Cargo Simulation

Physics-based digital twin of oil/chemical tanker cargo systems — inert gas, piping hydraulics, methanol dual-fuel, cargo heating, crude-oil washing, and intact stability. Deployed in production as a crew-training pilot.

LCSMar v0.2.0 →
OpsMar dashboard — weather stores, server fleet, containers, and GPU workstation status

On-Vessel AI

Local-first AI running at the edge on one GPU workstation aboard, over the ship's own corpus. Deterministic figures and page-level citations; nothing sent to an external cloud.

OpsMar →
QuantMar Research Studio — knowledge graph, signal editor, and LEAN falsification panel

Quantitative Research

Systematic equity research with falsification-first discipline: every hypothesis pre-registered, tested against a survivorship-free universe, and deflated for the number of strategies tried.

QuantMar →
WindMar full analysis — optimized route, voyage charts, weather profiles, and waypoint table
Featured

WindMar

Weather & Route Analysis

Combines hydrodynamic modelling, meteorological data, and constrained pathfinding to optimize routes accounting for hull resistance, propulsion efficiency, and cargo loading under realistic ocean conditions. A commercial-grade platform built and operated by a serving master mariner.

  • Global CMEMS/GFS weather layers with wave, wind, current, SST, and ice
  • Dual optimization: minimum fuel and minimum time with safety constraints
  • Full passage plan output with waypoints, ETA, and bunker estimates
  • Vessel models calibrated against measured consumption at sea
Open WindMar
LCSMar cargo synoptic — live loading simulation
Production

LCSMar

Liquid Cargo Operations Simulator · v0.2.0

A physics-based digital twin of a Reference MR Tanker's cargo plant, built from the vessel's class drawings. Cargo, ballast, inert gas, tank cleaning, heating, and methanol dual-fuel operations in one real-time engine. In production as a crew-training pilot for a shipowner.

  • Mass-conserving hydraulic network solver across five piping systems
  • Methanol bunkering, transfer, and engine-running with ESD logic
  • Stability and hull-girder strength validated against class documents
  • Loadicator-PDF condition library and instructor malfunction injection
Technical Documentation
OpsMar dashboard — weather stores, server fleet, containers, and GPU workstation
On-Vessel

OpsMar

On-Vessel AI Operations Hub

Local-first AI that runs at the edge on a single GPU workstation aboard the ship, over the vessel's own corpus — manuals, SMS procedures, noon reports, routes and calibration data. Questions are answered with deterministic figures and page-level citations; nothing leaves the ship.

  • Local-only LLM inference — no cloud dependency, works at sea
  • Page-level citations from the ship's own document library
  • Ship-performance engine (same physics as WindMar) for operational queries
  • CII, FuelEU Maritime, and EU ETS compliance reporting
  • Crew channel with cited risk assessments and knowledge graph
Explore OpsMar
QuantMar Research Studio — mechanism knowledge graph, signal authoring, and LEAN falsification
Research

QuantMar

Quantitative Research Studio

A research workbench for systematic equity strategies, designed around the conviction that most published backtests are overfitted. Runs the full hypothesis lifecycle: mechanism discovery from academic literature via knowledge graphs, formal pre-registration with hash-freeze, and deterministic falsification on untouched holdout data.

  • Knowledge-graph sourcing from 250M+ academic papers (OpenAlex)
  • Proprietary Python DSL — authored by LLM or human, sandbox-scored
  • Certified survivorship-free equity panel (1,400+ names, 22 years)
  • Deflated Sharpe ratio gate — honest about multiple testing
  • LEAN/QuantConnect oracle for independent validation
Research Publications

Contact

Research collaboration, technical discussion, or scoped advisory.

Get in Touch

committers.top badge