MyConstituency

We = DonGpt and myself

Inspiration

This project started with a form. A simple government form that asked for three things: my area MP, my local government, and my Traditional Authority (T/A). I didn't know any of them off the top of my head, and finding them turned into hours of digging through scattered PDFs and documents that were never meant to be searched.

Malawi — like a lot of developing countries — already goes without so much: reliable infrastructure, consistent public services, easy access to information that in other places is a quick search away. But technology doesn't have to be one more thing we go without. It's one of the few things we can fix without waiting on massive budgets or years of construction. If the data exists, it can be made accessible.

What it does

a lot but That historical view is one of the parts I'm most excited about — being able to see how a constituency has changed over time, and what patterns or possible outcomes emerge from that, is useful not just to ordinary citizens filling out forms, but to political parties and researchers trying to understand the electoral landscape. There's a lot more we want to layer on top of this foundation.

How we built it

Chat and I had to build our own extraction and cleaning pipeline entirely from what we could scrape out of the existing PDFs ourselves On the technical side, the original plan was straightforward: a Node.js/TypeScript backend using Hono, a React frontend, and JavaScript for the data-cleaning scripts. But with free OpenAI tokens in hand, we couldn't resist pushing further — we ended up building a Rust backend instead, betting on the performance upside.

Challenges we ran into

Government data wasn't built to be used. PDF-only records, no API, no bulk export, and a missing year of data we never managed to recover. No response from the MEC at any level we reached out to. Scope creep through over-engineering. Choosing Rust for the backend added real complexity — in hindsight, a simple REST API would have gotten us just as far, faster. Running out of tokens. Between data cleaning, debugging, and iterating on the backend, we burned through our OpenAI credits before we could finish everything on our list.

Accomplishments that we're proud of

I and Don GPT turned a pile of unstructured government PDFs into a working, queryable dataset — and built a real full-stack app around it.

What we learned

The power of friendship

What's next for MyConstituency

There's a lot more insight we want to add on top of the current map, history, and data-source views. The immediate blocker is simple: we're out of tokens. But once we're back up and running, the plan is to keep building — more features, more historical depth, and hopefully a tool that outlasts the hackathon and becomes something people in Malawi actually rely on.

Built With

• graphql
• gsap
• mec
• qwik
• rust
• supabase