EcoEPR: Tracking India's E-Waste from Collection to Recycler
EcoEPR connects India's e-waste collectors with authorised recyclers through intelligent item identification, transparent material valuation, verified pickups, and geotagged collection records. Built on AWS with Microsoft Foundry and Azure Speech, the platform demonstrates a practical approach to improving e-waste traceability and accountability.
EcoEPR: Proving Where India's E-Waste Goes
By Chirag Chaudhary · Team EcoEPR · WeMakeDevs × AWS Environmental Hacks · Track 03: Waste and Energy
India faces a growing e-waste challenge. In 2021–22, the country generated an estimated 1.6 million tonnes of e-waste, while approximately 0.52 million tonnes were collected and processed. A substantial share of discarded electronics also passes through informal collection networks. Meanwhile, brands must meet their regulatory recycling obligations through registered recyclers.
We built EcoEPR to connect these two worlds through traceable collection records, transparent material valuation, and a verified pickup workflow.
What EcoEPR Does
- Identify: Collectors upload a photograph and optionally record a voice note in Hindi. Azure Speech transcribes the audio, and Microsoft Foundry analyses the photograph and transcript together to identify the item and extract relevant fields with confidence scores.
- Value: The analysis identifies material types and estimates their masses. The server calculates the valuation using a predefined rate card, keeping pricing deterministic and auditable rather than relying on the model to generate the final price.
- Gatekeep: Every submission passes through a validation process before it can be accepted. Items can be approved, held for review, or rejected with a specific reason. For example, rule R-01 identifies submissions that are not e-waste.
- Coordinate pickups: Loads are routed to recyclers authorised for the relevant waste stream. The recycler verifies the pickup using a PIN from the collector's phone before a credit note is issued.
- Map: Geotagged intake records feed a hotspot map designed to show collection patterns across India's 735 districts, with ward-level detail in Mumbai.
Each material is priced against the server-side rate card, so the total can be traced back to individual line items.
Built on AWS
EcoEPR uses AWS for application hosting, identity management, data storage, and pickup coordination.
- AWS Amplify Hosting serves the web application.
- Amazon Cognito manages authentication and user roles.
- Amazon RDS for PostgreSQL stores the main application records.
- Amazon S3 stores uploaded photographs and voice notes, organised using SHA-256 hashes for content identification.
- Amazon API Gateway and AWS Lambda power the pickup workflow.
- Amazon DynamoDB stores pickup-related records, with DynamoDB Streams triggering the simulated UPI payout workflow.
For item identification and material analysis, EcoEPR uses Microsoft Foundry. Azure Speech handles Hindi voice transcription. AWS provides the surrounding application infrastructure, while these services power the implemented analysis workflow.
The architecture reflects the services actually used by the application.
What Fought Back
1. Amplify's 30-second execution ceiling
A model call taking 28.6 seconds returned a 504 error, while a 24.2-second call succeeded. We introduced a shared time budget for the request and return a readable timeout error at 26 seconds, rather than allowing the request to run into the platform's execution limit.
2. The 6 MB request-body limit
The request path through Lambda imposes a payload limit of approximately 6 MiB. Base64 encoding also increases attachment size by roughly one-third. We capped each attachment at 3 MB and added validation in both the browser and server to reject oversized files early.
3. Documenting the actual architecture
Our early write-ups mentioned Amazon Bedrock and Amazon Transcribe, but the application did not call those services. We corrected the documentation to identify Microsoft Foundry and Azure Speech as the services actually used.
This was an important lesson: a credible technical submission should describe the architecture that runs in production, not services that were considered during development but never integrated.
What We're Not Claiming
EcoEPR is a working prototype, not a claim that India's e-waste problem has been solved.
- Credits and UPI payouts are simulated.
- The recycler directory and rate cards contain sample data.
- Material masses are estimates generated through the analysis workflow, not laboratory assay results.
- The application coordinates and records e-waste collection; it does not physically process or recycle waste.
These limitations matter because traceability is only useful when the underlying records, verification steps, and real-world recycling outcomes can be trusted.
What's Next
Our next steps are to validate material estimates against real-world measurements, onboard verified recyclers, connect the workflow to genuine recycling documentation, and replace simulated credits with an appropriate production integration.
We also want to validate the district- and ward-level map against reliable geographic and collection data, helping stakeholders identify collection gaps and understand where recycling infrastructure may be needed.
Links and Resources
- Live application: EcoEPR — Prove it went to a licensed recycler
- Source code: https://github.com/chaudhary64/glowing-couscous
EcoEPR was built by Team EcoEPR for the WeMakeDevs × AWS 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