Flag It.

Inspiration

Imagine you're walking in the park and you see a glass shattered on the floor. What would you do? Probably what most people would do, walk past it. Not because you're a bad person. Not because you don't want to help. But because you don't know what you can do. Who to contact, what to say. Well, what if all you had to do was take a picture and press send? The right people get informed and the problem gets solved! Well, that in a nutshell is Flag It.

What it does

Flag It. allows you to capture hazards from your camera and submit a report in seconds, sending the what3words location, description and photo of the hazard to the relevant authority.

How we built it

We used Claude to analyse and annotate the image, filtering out images with no hazards, and used another Claude agent to iteratively find the contact information for the most appropriate authority. For example, for litter in the SCR, Flag It. would inform the Imperial Estate Management team by email; for a pothole in the road, Flag It. would alert the local council.

Challenges we ran into

Evaluating the urgency of a certain hazard proved a challenge. Finding the best point of contact for each report was also difficult.

Accomplishments that we're proud of

We implemented a feature where reports similar in description, location and time were grouped together and reported as one. Fine-tuning our final decision equation proved technically difficult as we had to weigh independent factors such as time, location and urgency against one another.

What we learned

Assigning tasks to AI Agents, designing a simple, well thought-out UI/UX, fine-tuning similarity metrics, working with Node.js for mobile applications, server-side optimisation.