07 Hackathon project with Dhruv & Friends · 2026
BlueNet: Ocean Watch
A prototype that helps people protecting Mauritius's ocean spot suspicious fishing boats and share wildlife sightings from the water.
A walkthrough of the prototype: boats that switch off their tracking appear in a ranked list and on a map, the AI assistant writes a case file, and a phone app reports sightings even offline.
How it works
Rules flag the boats worth a look; an AI assistant prepares the case file.
Gather signals
Boat tracking data, protected areas and tips from people at sea.
Map dataScore
Fixed rules give each event a clear score with reasons.
PythonReview
A dashboard lists the most suspicious boats on a map.
ReactInvestigate
An AI assistant on the device checks the boat's background and writes a case file.
Gemma AIClear rules decide which boats look most suspicious; the dashboard shows why; an AI assistant on the device gathers background and writes a case file for a person to review.
01 Problem
Why this exists
Mauritius has a huge ocean territory. Signs of suspicious boat behaviour and first-hand wildlife sightings often sit in different places, or are never recorded at all. We explored how one tool could bring both together.
The same Dhruv & Friends team built BlueNet during the 24-hour Build with Gemma: Blue Mauritius & Beyond hackathon, where it received the Coup de Cœur award. We tidied it up afterwards.
02 My part
What I built
Built the dashboard: a ranked list of suspicious boats, a map, filters and a detail view for each boat.
Built an AI assistant that runs on the device, checks a boat's owner, sanctions, port history and fishing licence, then writes a case file.
Added Kreol support so tips from the water can be understood and translated privately.
Trained a small Kreol language add-on using reviewed examples.
Prepared the hackathon submission and documentation.
03 Approach
How I approached my part
Flag, don't accuse
A high score means 'worth a look', not proof of wrongdoing.
Rules rank, AI researches
Fixed rules decide the order; the AI only gathers background and drafts the case file.
Built for Mauritian Kreol
Keep people's original words and improve how the system understands local terms.
Grow with the community
Let skippers, tour operators and divers share useful sightings from the water.
04 Result
Where it is now
A working prototype with a scoring system, my investigation dashboard and AI assistant, and a phone app for reporting sightings, even without signal. It received the Coup de Cœur award.
BlueNet is a hackathon prototype, not a real surveillance system. Some boats and records in the demo are sample data, case files need a person to check them, and the Kreol support still needs testing with more speakers.
Next: Grow it into a privacy-aware sightings network for skippers, tour operators and divers. For dolphin tourism, shared approximate sighting areas could help crews without revealing exact animal locations.