I've won four hackathons — 1st at Laksh 2026, 1st at Prompt War, 2nd at Brand Combat, 3rd at Rising India. These are the project ideas and strategies behind those wins, written for anyone about to spend 24–48 hours building something with AI.
An AI tool that maps a student's skills and interests to realistic career paths, then generates a step-by-step roadmap. Judges love it because the problem is universal and the demo is instant: type in your background, watch a personalised path appear. I built a version of this as one of my own projects — it works because everyone in the room has felt the confusion it solves.
An app that finds nearby EV charging stations, checks availability, and plans routes around charging stops. It mixes maps, real-time data and a clean mobile UI — a strong combo for hackathons because it demos well on a phone and solves a problem that is visibly growing. This was one of my own builds, and the live map demo always lands.
A game where players write prompts to make an AI produce a target output — judged by similarity scores. It turns prompt engineering into a spectator sport, which makes it memorable on stage. My 1st place at Prompt War came from understanding exactly this: how small wording changes move a model's output.
Feed it a product, get back a full brand pitch: name, tagline, positioning and a 30-second script. My 2nd place at Brand Combat taught me that branding challenges are really about structured creativity — and AI is very good at structured creativity when you constrain it well.
My 3rd place at Rising India and 1st at Laksh 2026 both came from the same playbook: pick a problem people around you actually face, then apply AI narrowly and well. A tool that helps local shops, students or commuters beats a generic chatbot every time, because judges can verify the need is real.
You have minutes on stage. Build the one flow that makes people say 'oh, that's useful' and make it flawless. Everything else can be a mock.
'AI for students' loses. 'AI that turns your syllabus into a revision plan' wins. Specificity is what makes a judge believe you'll finish.
Most teams call the model once and hope. Structured prompts, few-shot examples and output validation are free performance that most competitors skip.
A memorable name and a one-line pitch carry further than another slide of architecture. Judges remember 'CodePulse', not 'our AI-powered solution'.