Sanit.AI

Inspiration

Across India and globally, sanitation is measured by access, not by effectiveness. Infrastructure is declared complete even when water is unsafe, drains fail, and contamination spreads quietly. Communities lack visibility, verification, and a way to demand action. Sanit.AI was inspired by this gap between presence and performance, and by the belief that clean water is a basic human requirement, not a privilege.

What it does

Sanit.AI is an AI driven sanitation and water risk analysis platform that works across all devices through the web. Users upload images of sanitation infrastructure such as drains, handpumps, tanks, or filters. The AI analyzes visible hygiene and structural risks, explains issues in simple terms across 10+ Indian languages, suggests low cost fixes, provides groundwater context, and enables AI assisted Public Interest Litigation drafting to push authorities to act.

How We Built It

We built Sanit.AI as a modern web application using Next.js and React, deployed on Vercel's cloud infrastructure for reliable global access. The platform integrates advanced AI vision models through Groq's API to analyze sanitation images, with intelligent retry logic to ensure consistent service even during high demand. Images are securely processed and validated before being analyzed, with all results stored in a PostgreSQL database managed through Supabase. The frontend is optimized for performance with minimal JavaScript loading, smart caching strategies, and a component-based architecture that works seamlessly across phones, tablets, and computers. We implemented multilingual support for 12 Indian languages with dynamic content switching and built-in accessibility features.

The groundwater analysis system uses real hydrological formulas to calculate water sustainability metrics, processing authenticated data from monitoring stations across all Indian states. It evaluates factors like aquifer storage, rainfall recharge rates, and extraction patterns to predict depletion timelines and provide actionable recommendations. We integrated comprehensive water quality datasets covering nearly 200 monitoring locations nationwide, cross-referenced with WHO and BIS safety standards to assess contamination risks. The AI system receives contextual information about the user's region and specific concerns, generating detailed analysis in their preferred language while maintaining structured data for reliable processing. Security measures including encrypted connections, input validation, and database access controls ensure user data protection throughout the platform. Live platform: https://v0-sanitation-image-analysis.vercel.app/

Challenges we ran into

Sanitation failures vary widely in appearance and image quality. Chemical contamination cannot be detected visually, and water quality data is often incomplete or outdated. Ensuring consistent performance across devices and networks was also challenging. We addressed this by positioning the platform as an early warning and decision support system rather than a lab replacement, and by optimizing the platform for lightweight web delivery. Additionally, we can optimize through more user usage, allowing for more data inputted into the algorithms we have to ensure the algorithm can progressively get more accurate

Accomplishments that we're proud of

We built a scalable, cross platform AI sanitation inspection system accessible to anyone with internet access. We integrated AI assisted PIL generation to connect community evidence with legal accountability. The platform aligns directly with WHO recommended visual sanitary inspections and SDG 6 targets, turning an accepted but manual process into an automated, scalable solution. Especially the fact it can help empower our target audience of rural indians. Additionally, we validated our AI-PIL(public interest litigation) generation by a real Lawyer, who stated that it was really accurate to what a real submitted PIL would be like.

What we learned

Accuracy alone is not impact. Accessibility, language support, simplicity, and legal empowerment are critical for real world adoption. We also learned that institutions do not need new systems, they need structured information in formats they already accept. Transparency about AI limitations builds trust with users.

What's next for Sanit.AI

We plan to expand and fine tune the dataset, improve groundwater risk coverage in vulnerable regions, and strengthen community reporting through NGO partnerships. We will integrate low cost edge devices for basic real time water sensing while maintaining universal web access. Our goal is to scale Sanit.AI into a nationwide early warning and accountability layer for water and sanitation.

Built With

• cva
• datefns4
• groqsdk
• groundwater36
• lucidereact
• nextjs16
• postgresql
• radix
• react19
• reacthookform7
• recharts2
• shadcnui
• sonner
• sql
• supabasessr
• tailwind4
• typescript5
• vercel
• vercelai
• watercsv
• webspeechapi
• zod3