PrakratiAI

Inspiration

The constant rise in electricity bills, inefficient energy usage at home, and a lack of daily waste segregation awareness are real-world problems faced by almost every household. While many people want to reduce their carbon footprint or save money on utility bills, they often don't know where to start or which specific appliances are draining the most power. We wanted to build an all-in-one platform that bridges the gap between environmental sustainability and financial savings—an intelligent assistant that translates complex environmental data into simple, practical, and actionable steps. That vision led to the creation of Prakriti AI.

What it does

Prakriti AI is a comprehensive Sustainability & Energy Audit Platform designed to help users make eco-friendly and cost-effective decisions: 1->Electricity & Water Audits: Users input appliance details or monthly bill amounts, and the platform delivers a structured Consumption Breakdown, pinpoints the Reasons for High Bills ("Why"), and provides a Step-by-Step Reduction Strategy ("How"). 2-> Multimodal Waste Classifier: Users can upload a photo of garbage or electronic items, and the system classifies the item into Recyclable, Organic, or E-Waste/Hazardous, complete with proper disposal guidelines. 3-> Eco-Route Planner: A dual-input interface (Origin & Destination) that compares travel modes based on Time, Estimated Cost, and CO₂ Emissions. 4-> Interactive Solar ROI & Appliance Calculators: Custom tools to estimate solar rooftop feasibility, payback periods, and track daily sustainable habits. 5-> Sign-Out Eco Audit Summary: Automatically generates a comprehensive session summary report upon sign-out that users can view or export as a PDF.

How we built it

Prakriti AI was engineered using a modern, fast, and serverless stack: 1->Frontend Framework: Built with React.js and Vite for ultra-fast load times and smooth rendering. 2->Styling & UI: Tailwind CSS, Glassmorphic UI overlays, Lucide React icons, and a custom background grid. 3->Database & Authentication: Powered by Supabase (PostgreSQL) for persistent chat history, user sessions, and strict Row Level Security (RLS) policies. 4->AI Engine: Integrated Google Generative AI (Gemini Flash) for multimodal image classification and domain-specific prompt engineering. 5->Markdown Processing: Dynamic Markdown parsing to display formatted tables, bold metric highlights, and clean audit reports.

Challenges we ran into

Building Prakriti AI involved solving several real-world engineering hurdles: 1->API Rate Limits & Quota Overflows: Unrestricted grounding calls caused 429 rate limits. We restructured system prompts and engineered knowledge-retrieval fallbacks to maintain a 100% stable response stream. 2->Payload Formatting Errors: The SDK required text inputs to be formatted strictly as object arrays. We implemented client-side map-serialization to ensure clean input payloads. 3->Dynamic UI Mode Switching: Designing a smooth UX transition between a standard single-prompt input and a dual-input layout (@route mode: Origin + Destination) required careful state management. 4->Scope Enforcements & Hallucinations: To prevent the AI from answering off-topic queries (like general coding trivia or politics), we implemented strict Domain Guardrails to keep responses strictly focused on sustainability and energy audits.

Accomplishments that we're proud of

1->Shipped a fully functional Serverless Full-Stack Web Application with fast response times. 2->Successfully merged real-world financial metrics (₹ savings) with environmental metrics (kg CO₂ reduced) in a unified UI. 3->Built a working multimodal vision pipeline for image-based waste classification. 4->Secured user database records with granular Row Level Security (RLS) policies.

What we learned

1->Prompt Engineering is Context Engineering: We learned how critical domain guardrails and structured system instructions are for obtaining reliable outputs from LLMs. 2->BaaS Architectural Efficiency: Experienced how seamlessly Supabase and React can replace traditional backend servers for full-stack deployment. 3->Edge-case Management: Understood the importance of error boundaries, payload validation, and image optimization in production-ready apps.

What's next for PrakratiAI

1->Native Mobile App: Developing a React Native application for direct camera access during waste classification. 2->IoT Smart Meter Syncing: Integrating direct API syncing with smart meters for real-time energy spike detection. 3->Local E-Waste Collection Drive: Connecting users directly with local recycling NGOs and municipal collection points. 4->Gamified Community Leaderboard: Introducing eco-badges, streak tracking, and community leaderboards to drive long-term user engagement.