Projects gallery
Twelve real projects from past cohorts — each one built on a genuine marine dataset, most of them debugged into existence during a high tide nobody could walk away from.
Overlays daily wind direction onto recorded litter pickups to guess where the next wave of plastic will beach. Tuned the wind-weighting during a six-hour high tide until it actually predicted Tuesday’s mess.
Counted limpet movement per tide cycle from time-lapse frames, then trained a tiny classifier to flag a moving shell. The causeway flooded mid-debug — the only reason the loop finally got fixed.
Scored underwater noise against the ferry timetable and found the quietest hour for porpoise listening. Presented as a chart the harbormaster now actually uses.
A terminal widget that says exactly how many safe minutes remain on the causeway, with an amber bar that drains as the water rises. Now runs on the mess-hall screen.
Segmented seagrass beds from drone imagery across three summers and measured the meadow actually growing. The first project to get cited in a real survey report.
Separated weather surge from astronomical tide in the island gauge record and found the causeway floods eleven minutes early on a falling barometer.
Counted green crabs crossing the slipway per tide phase with a motion-diff script. Concluded crabs also prefer the low-tide window, which felt validating.
Sonified two weeks of salinity readings into a melody where the estuary’s freshwater pulse is audible as a falling phrase. Demo day had headphones.
Mapped where each tide left its seaweed line and reconstructed the month’s tide heights from photos alone — a tide gauge made of wrack.
Digitized four summers of paper sighting logs and charted puffin arrivals against sea temperature. Built the data-entry form her nan now uses.
Flagged snagged fishing-gear candidates in sonar imagery of the channel, ranked by confidence. Two of his top five were real nets, since recovered.
Showed each tide pool has its own microclimate that resets with every flood — plotted as small multiples pinned above the lab bench like postcards.