AI-powered ocean early warning network

Background

Marine ecosystems around the world are losing life to human activity that nobody is watching in real time: ghost gear (lost or abandoned fishing nets) that keeps entangling animals, bycatch, and collisions between vessels and marine animals.

Today's tracking relies on manual logs, isolated sensors and local hardware. These don't scale, and they can't keep up with ocean conditions that change by the hour.

The real gap

Human activity and marine life movement are modelled separately, not as one connected, changing system. Shipping knows where ships are going; researchers know where turtles and dolphins tend to be. The two pictures rarely meet in time to change a decision.

As a result, millions of non-target animals, including dolphins and sea turtles, are lost every year to encounters that could have been prevented.

Who you are designing for

The people who make decisions at sea: shipping companies, fishing fleets, ferry and port operators, and the marine agencies and conservation groups who work with them. The animals can't use your product, but every design decision is ultimately for them.

What your concept should cover

1. The signals. What data would you combine? Think vessel positions, sightings, migration seasons, sea temperature, currents, fishing activity.
2. The forecast. What would the AI predict, how far ahead, and how sure would it need to be?
3. The warning. Who gets told, through what channel, and what exactly are they asked to do?
4. The routing. How does the network suggest, or automatically make, a safer route or a change of timing?
5. Why the cloud. Oceans cross borders. How does your network work at global scale and in real time?

Things to think about

• The cost of a detour. Rerouting costs fuel and time. Why would a captain follow your warning?
• Connectivity at sea. What happens when a vessel is offline for hours?
• False alarms. Too many warnings and people will start ignoring them.
• Who pays, and who shares data. Shipping companies, governments and researchers all have different incentives.

What to deliver

Problem research

Who is affected, how often, what they do today and why it fails. Cite your sources.

A user and their journey

One specific person, before and after your solution, as a storyboard or journey map.

Concept architecture

A diagram of what data goes in, what the AI does, and how it all connects. No code required.

Mock-ups

Screens, a no-code prototype or a physical sketch, so the judges can picture it.

A 10-minute pitch

Including impact, feasibility and risks. Every member ready for Q&A.

Qwen is a strong research partner: use it to explore the problem, then check what it tells you against real sources.

How it's judged

Two judges score every team on this problem statement out of 100, using the non-technical track rubric.

1. Problem insight25 pts

   Real research: who exactly is affected, how often, why it matters, and what they do today.

2. Solution & user journey25 pts

   A clear concept someone can picture, shown through mock-ups, a storyboard, a prototype or a user flow.

3. Technology & AI reasoning20 pts

   Why AI and technology suit this problem, what they contribute, and roughly how information would flow through the solution.

4. Feasibility & impact15 pts

   Honest about risks, cost, implementation and adoption, and clear about what measurable success looks like.

5. Pitch15 pts

   Persuasive, structured, on time, and confident under questions.

Judges reward appropriate use of AI, not complexity for its own sake. Using Alibaba Cloud, Qwen or Qoder is supported but not required to score well.

Draft · final weights are confirmed before the workshops. Compare both track rubrics