ShieldBin @ Environmental Hacks by WeMakeDevs x AWS
Autonomous Source-Level Waste & Contamination Intelligence Inspector
ShieldBin: Smarter Waste Decisions, Powered by AI ♻️
Turning AI into action for a cleaner, more sustainable future.
Team Sankalp Squad | Bharat Builds Tour — Environmental Hacks, Track 03: Waste and Energy
The Problem: Small Mistakes, Big Environmental Impact
A greasy pizza box, a discarded battery, or a misidentified plastic container—one wrong disposal decision can contaminate an entire recycling stream.
The problem isn't just a lack of awareness; it's the lack of instant, reliable guidance at the moment people need it most.
What if AI could help people make smarter waste-disposal decisions before waste even reaches the bin?
That question inspired us to build ShieldBin.
Meet ShieldBin: Intelligence at the Source
ShieldBin is an AI-powered waste inspection and contamination intelligence platform designed to make waste segregation smarter, simpler, and more accessible.
It brings together four key capabilities:
- AI-powered inspection to analyze visual information about waste.
- Policy-based evaluation to assess disposal decisions against configured rules.
- AI Copilot to answer waste-related questions conversationally.
- Tracking and gamification to encourage responsible segregation habits.
How It Works: From Image to Insight and Architecture
ShieldBin combines AI-driven analysis with structured policy evaluation to create a more informed disposal workflow.
Visual Input → AI Analysis → Policy Evaluation → Disposal Guidance
Our architecture integrates Google Gemini API, Amazon Bedrock, and the AWS Cedar policy engine, with policy references covering Indian waste-management guidance. This approach separates AI interpretation from rule-based evaluation, helping make recommendations more structured and explainable.
Use of AWS :
AWS Open Source Stack =
AWS Cedar (cedar-policy) - Used as a Policy-as-Code engine to enforce Indian environmental laws (MoEFCC SWM 2016 & CPCB E-Waste Rules). Evaluates disposal actions into deterministic PERMIT or FORBID legal verdicts with official citations. Crucially, whether Google Gemini or Amazon Bedrock perceives the image, AWS Cedar remains the sole deterministic authority deciding statutory compliance (
PERMIT vs FORBID ). The LLM perceives, Cedar decides.Boto3 & Botocore - The AWS SDK for Python used by the backend to connect, authenticate (SigV4), and invoke Amazon Bedrock and Amazon DynamoDB with built-in retry handling.
AWS Services =
Amazon Bedrock (bedrock-runtime) - "Integrated via Boto3 as our enterprise foundation model fallback. For live camera streams, real-time edge responsiveness is critical—so active continuous frame analysis runs through Google Gemini Flash (~500ms latency), while Amazon Bedrock (Claude 3.5 Sonnet / Haiku) provides deep audit fallback for enterprise AWS environments."
Amazon DynamoDB - Serverless NoSQL database managing two tables:
ShieldBin_Scans: Audit log storing scan results, items detected, and Cedar legal decisions.
ShieldBin_Scans_Scores: Live ledger tracking household eco-points, contamination prevented, and ward-level accuracy rates.
Your Waste Questions, Answered
Not sure whether a greasy pizza box is recyclable? Need clarification about an inspected item?
The ShieldBin AI Copilot lets users ask questions, explore disposal considerations, and get conversational assistance—all within the same experience along with voice access.
Combined with scan tracking and gamification features, ShieldBin aims to make responsible waste segregation easier to understand and build into everyday habits.
Built with Modern Cloud and AI Technologies
Our technology stack combines:
- Frontend: React, TypeScript, Vite, Tailwind CSS
- Backend: Python, FastAPI
- AI: Google Gemini Multimodal Vision API (Sub-second Live Stream Engine) & Amazon Bedrock (Claude 3.5 Sonnet / Haiku Enterprise Foundation Model)
- Policy Engine: AWS Cedar
- Data Layer: Amazon DynamoDB
- Deployment: Vercel and Render
What We Learned—and What's Next
Building ShieldBin reinforced one important lesson: AI predictions alone aren't enough. Real-world solutions need structured rules, intuitive user experiences, and honest acknowledgment of uncertainty.
Our next steps include improving inspection reliability, expanding waste-category coverage, strengthening location-specific guidance, and testing recommendations against real-world scenarios.
A Cleaner Future Starts with One Better Decision
ShieldBin isn't about claiming AI can solve waste management on its own. It's about using technology to make better decisions easier—right where waste is generated.
Less confusion. Smarter segregation. A cleaner tomorrow.
We're Team Sankalp Squad, and we're building toward a future where responsible disposal starts with informed action.
Try ShieldBin
Blog written by - Garv Ghildiyal
Meet Team Sankalp Squad
- Garv Ghildiyal - Team Leader ( Tester and Reviewer )
- Abhinav Singh (Demo video and Pitch)
- Devesh Singh Rawat ( Backend and AI Designer)
- Abhinav Chauhan ( Frontend and UI/UX Desinger
Built for the Bharat Builds Tour — Environmental Hacks, Track 03: Waste and Energy.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article