When I joined the PromptWars Virtual hackathon, I knew I wanted to build something that mattered. The Indian election process is massive and, honestly, pretty daunting if you're a first-time voter. I wanted to build a bridge—a portal that makes understanding your democratic rights as easy as sending a text.
That’s how VoteGuide India was born: an interactive guide with visual timelines, maps, and an AI chatbot to answer any burning questions.
But building all of this alone during a hackathon? That’s where AI came in. I didn't just plug an AI API into the app; I used AI as my actual development partner. Here’s how it went down.
My Toolbelt & Why I Chose It For the core AI feature (the chatbot), I went with Google Gemini 1.5 Flash. Why? Because when you're dealing with elections, speed and safety are non-negotiable. I needed a model that was snappy but wouldn't start spouting political biases or hallucinated opinions.
Behind the scenes, I used AI coding assistants constantly. I leaned on them to help scaffold my React components, figure out weird CSS flexbox issues, and set up my Express server. It was like having a senior dev sitting next to me whenever I got stuck.
The Evolution of a Prompt If there’s one thing I learned, it’s that your first prompt is never your last.
When I first hooked up the chatbot, my prompt was basically: "You are an election assistant. Answer questions about voting." The result? It was okay, but it gave generic answers and sometimes pulled in rules from US elections!
I tweaked it to be more specific: "You are an Indian election assistant. Answer based on the ECI." This was much better, but if I pushed it, it would still sometimes playfully offer an opinion.
Finally, I locked it down with a much stricter constraint: "You are an official, politically neutral guide for the Indian electoral process. You must only provide factual information based on Election Commission of India (ECI) guidelines. Under no circumstances should you endorse a candidate, party, or offer political opinions. Format your answers clearly." Bingo. It instantly sounded like an official, helpful, and totally neutral government guide.
The Human-AI Division of Labor The coolest part of this hackathon was realizing what AI is good at versus what I’m good at.
I let the GenAI handle the tedious stuff: generating boilerplate code, figuring out why my API routes were crashing, and, of course, dynamically answering user questions in the app itself.
This freed me up—the human—to focus on the stuff that requires empathy and taste. I spent my time designing a clean, Government of India-inspired UI. I made sure the app was fully accessible (WCAG 2.1 AA) so everyone could use it, and I meticulously curated the Google Maps integration to show the exact locations of election offices.
Wrapping Up Building VoteGuide India was a huge learning curve, but it proved to me that AI isn't here to write our apps for us—it's here to take away the friction so we can focus on building things that actually matter.
Top comments (0)