Articles
METHODS / SYSTEMS / IMPACT
Technical notes on deep learning, reinforcement learning, world models, maritime intelligence, fisheries monitoring and responsible AI. I write from the model outwards: assumptions, data, evaluation, operational impact and failure modes.
From Fishnet to Meenu: from detector accuracy to reviewable fisheries evidence
What changes when fisheries computer vision moves from an open dataset to event-level evidence, cross-vessel evaluation and human review.
Prediction is not intervention: a causal view of port-call optimisation
Why port-call optimisation should move from ETA prediction to causal intervention modelling over vessel, berth, weather, cargo and stakeholder decisions.
Maritime intelligence for commodity flows: from vessel tracks to probabilistic evidence
How vessel, port and corridor events can become a time-consistent evidence layer for commodity-flow forecasting without pretending AIS reveals cargo directly.
What commercial fisheries need from electronic monitoring
A commercial EM product needs evidence capture, review, traceability, standards mapping and measurable workflow value—not a dashboard around a detector.
Species signatures and a sampling doctrine for fisheries AI
Representation learning and sampling policy should be designed together so species models generalise across vessels and review resources are spent where they change evidence quality.
Fish detection is not enough: evidence for ETP interactions
ETP monitoring needs event reconstruction, rare-event evaluation, independent audit coverage and evidence provenance—not only species accuracy.
Electronic monitoring data requirements: what to collect, why it matters, and what AI changes
A system-level view of EM data: cameras, sensors, event schemas, retention, review design and the metrics needed to show real monitoring impact.
Observer data and eLogbooks: the evidence layer before AI
Why observer records and eLogbooks remain the backbone of fisheries evidence, and how to design them for validation, traceability and AI-assisted review.
Transshipment detection should model an interaction, not two nearby tracks
From proximity thresholds to probabilistic interaction modelling for fishing-vessel and carrier encounters.
Dark shipping: absence of AIS is evidence, not a verdict
A dark-vessel system should model expected AIS reception, multi-sensor detections, identity consistency and competing hypotheses.
Why a 99% maritime anomaly score may tell you very little
A practical evaluation checklist for AIS anomaly models: label provenance, synthetic events, shared test sets, leakage, class definitions and independent evidence.
Loitering, waiting and anchorage: define the state before training the detector
A literature-driven argument for separating loitering, waiting, drifting and anchorage before building a maritime anomaly detector.
Anchorage detection is a behaviour-inference problem
Why anchorage detection should be framed as latent-state inference over imperfect AIS rather than a speed threshold inside a polygon.