Prism
One URL. Three ways to understand it.
Prism was designed and built by Yadnyesh Mulay, an AI-first full-stack developer based in Nashik, India. Paste a website. Pick a mode. Get a fast TL;DR, a full song written from the page, or a kid-friendly explanation with a story and a quiz, one mode at a time, real AI under the hood. Built for the Outskill Hackathon.
The problem
Web content is often too long, too dense, or too technical. People need different entry points: a quick summary, an engaging format, or a simplified explanation.
The approach
A single URL input that routes to three distinct AI transformations, each with its own prompt engineering, output formatting, and UX considerations.
How it was built
Prism transforms any webpage into three distinct formats: a concise TL;DR summary, an AI-generated song with lyrics (and optional audio via ElevenLabs), or an ELI5 explanation complete with a story, fun facts, and an interactive quiz. The project demonstrates end-to-end AI integration: server-side content extraction, multi-modal LLM prompting (Gemini), and graceful fallback handling for paid API features.
The architecture uses a custom "Clay" design system built with CSS custom properties and CSS Modules, no utility framework, implementing a warm, light interface with a six-color feature card palette. The frontend is a Vite + React SPA with a sophisticated state machine managing the analysis pipeline. Backend runs on Vercel/Netlify serverless functions with a 60-second timeout ceiling that comfortably handles the 12-17s song generation pipeline.
Where the AI sits
Core product, multi-modal LLM pipeline with three distinct prompt strategies (summarization, creative lyrics generation, pedagogical explanation + quiz generation). Real AI, not a wrapper.
Stack
- React
- Vite
- TypeScript
- Google Gemini API
- ElevenLabs Music API
- Cheerio
- Vercel/Netlify Functions
- Custom CSS Design System
Links
More work by Yadnyesh Mulay: all projects · about Yadnyesh Mulay · home