A Real-Time Task Optimizer at IIT Bombay

This was our entry for the HERE hackathon at IIT Bombay's TechFest, built by team "5G Only." It placed in the top 10 internationally, which for a hackathon project I'm still happy about. The problem we picked was one everyone recognizes: you have a handful of errands to run, and the order you do them in quietly decides how much of your day you waste in traffic.

Why I made it

Most to-do apps treat tasks as a flat list. But errands happen in the physical world — a bank, a pharmacy, a friend's place — and the "best" order depends on where you are and what the roads look like right now. I wanted an app that would take a list of things to do and figure out the smart sequence for me, factoring in live location and traffic instead of just check-boxes.

How it was built

It's a Flutter app that leans on the HERE platform:

• here_sdk for maps and routing,
• geolocator to track the user's current position,
• shared_preferences for local persistence,
• and a small Python backend that talked to HERE's places API and layered generative AI on top of the results.

The app is organized around a few clear pages: a home view, a task list, a map/route view, and a profile. The backend did the enrichment work — pulling nearby places, trimming each result down to the fields we actually cared about (title, category, distance), and shaping that into something the app could use to reason about a route.

The generative-AI piece was what made it more than a router: rather than only spitting out coordinates, it helped turn the raw location and traffic data into a sensible, human-readable plan for the sequence of tasks.

What I learned

• Real-world constraints make a demo click. "Reorder my errands by traffic" is instantly understandable to a judge in a way that abstract optimization isn't.
• Location APIs are deep. HERE's SDK does a lot, and most of our time went into getting permissions, positioning, and routing to behave on a real device.
• Small backends punch above their weight. A thin Python service doing places lookups and reshaping JSON was enough to make the app feel intelligent.

Like most hackathon builds it's rough around the edges, but the core idea — that your task list should respond to the world around you — is one I still think is right.

If you want to look at the code, here it is: HereHacathonIITB on GitHub.