AWS Builder Center

Builder Center

RE:GENESIS: Closing the Loop on Enterprise E-Waste with AWS

An AWS-powered e-waste recovery pipeline that uses computer vision, BOM auditing, reuse-value ranking, and KMS-signed Second-Life Passports to help recover reusable enterprise hardware instead of shredding it.

RE:GENESIS: Closing the Loop on Enterprise E-Waste with AWS

Team: RE Event: Environmental Hacks — Bharat Builds Tour Track: Waste and Energy GitHub:  mitra9917/ReGenesis  Live Demo:  Launch RE 

The Problem: Recover Components, Not Just Metal

Enterprise servers, laptops, and networking equipment often contain reusable components such as GPUs, RAM, SSDs, and other valuable hardware. However, when devices reach end of life, shred-first recycling can destroy functional components and waste the energy and resources invested in manufacturing them.
Recovered hardware also needs trustworthy provenance. Buyers and reuse marketplaces need evidence that a component came from a known device and that its recovery record has not been tampered with.
RE addresses this challenge by making component recovery measurable, auditable, and verifiable. It helps operators identify reusable parts, prioritize disassembly, and generate cryptographically signed records for second-life hardware.

Our Solution

RE is an AWS-powered, adaptive electronics recovery platform designed for IT Asset Disposition (ITAD) operators, circular-hardware recyclers, and hardware reuse marketplaces.
An operator uploads an image of a device's internal components or selects a known device model, such as a Dell PowerEdge R740, ThinkPad T14, or Cisco Catalyst 9300.
The recovery pipeline then follows six stages:
  1. Detect: Computer vision uses a YOLO model hosted on Amazon SageMaker to identify hardware components. Catalog-assisted fallback helps the workflow continue when visual detection is uncertain.
  1. Audit: Detected components are compared against the device's Bill of Materials (BOM) to identify expected, detected, and potentially missing components.
  1. Plan: A Reuse Value Score (RVS) ranks components to help prioritize disassembly based on reuse potential.
  1. Test: Simulated diagnostics assess component status as part of the recovery workflow.
  1. Passport: AWS KMS signs Second-Life Passports, allowing signatures to be verified and tampering to be detected.
  1. Impact: The platform estimates material diverted from shredding and potential CO₂-equivalent emissions avoided using ranges rather than unsupported precision.
The goal is to move from shred-first disposal toward component reuse supported by traceable evidence.

Architecture and Technology Stack

The frontend is built with React and Vite and deployed as a static website on Amazon S3. The backend is a serverless recovery pipeline deployed using AWS Serverless Application Model (AWS SAM).
The workflow follows this sequence:
Ingest → Detect → Audit/Parse → Plan → Test → Passport → Impact

AWS Services Used

  • Amazon API Gateway: Exposes REST API endpoints to the frontend for ingestion, recovery jobs, and passport verification.
  • AWS Lambda: Executes the individual stages of the recovery pipeline.
  • AWS Step Functions: Orchestrates the RecoveryFlow and provides a structured, auditable workflow.
  • Amazon S3: Stores uploaded images, logs, passport artifacts, and the static frontend.
  • Amazon DynamoDB: Stores device records, components, passports, and pipeline events.
  • Amazon SageMaker: Hosts the YOLO endpoint used for component detection.
  • AWS Key Management Service (AWS KMS): Supports signing and verification of Second-Life Passports.
  • Amazon CloudWatch: Provides logs and operational visibility.
  • Amazon SNS: Supports job-completion notifications.
  • AWS SAM CLI: Defines and deploys the serverless application through infrastructure as code.
Each service has a specific role: Step Functions orchestrates the workflow, SageMaker provides computer vision, KMS supports cryptographic trust, and S3 and DynamoDB preserve the associated evidence.

Design Decisions and Reliability

Catalog-Assisted Fallback

Real-world device images vary in angle, lighting, and visibility. A vision-only workflow could fail when detections are uncertain. RE therefore supports catalog-assisted fallback so that the recovery workflow can continue when the vision model cannot confidently identify components.

Cost-Aware Computer Vision

Amazon SageMaker endpoints can incur costs while running. We added an explicit endpoint status and on/off control to support cost-aware demonstrations and on-demand vision inference. This also highlighted the need to account for endpoint startup and waiting time in the user experience.

Verifiable Second-Life Passports

A conventional PDF certificate alone does not establish tamper-evident provenance. RE uses AWS KMS signing and verification to provide a cryptographic verification flow. The demonstration also includes a tamper-rejection path to show what happens when a passport's integrity check fails.

Impact: Making Reuse Measurable

The Impact view is designed to communicate the environmental value of component recovery through:
  • Estimated kilograms of material diverted from shred-first disposal.
  • Estimated CO₂-equivalent emissions avoided, expressed as ranges.
  • A framework for thinking about recovery across fleets of devices rather than a single chassis.
We intentionally avoid presenting unsupported environmental figures as measured results. The objective is to make reuse decisions more transparent and provide a foundation for evaluating recovery at scale.

What We Learned

1. Cloud inference requires cost awareness. Managed machine-learning endpoints introduce both operating costs and startup latency. Cost controls and fallback paths help keep a demonstration practical.
2. Real-world vision needs a resilient fallback. Component images are not always clear. Combining model-based detection with catalog-assisted workflows makes the overall recovery process less dependent on perfect inference.
3. Trust must be built into the product. Second-life hardware needs verifiable records, not just a certificate. Integrating signing and verification into the recovery workflow makes provenance a core feature.
4. Infrastructure should support auditability. Using Step Functions to orchestrate distinct recovery stages makes the pipeline easier to inspect, understand, and demonstrate.

Try RE

  • Architecture Documentation: See  docs/ARCHITECTURE.md  in the repository.
Built by Team RE for Environmental Hacks — Bharat Builds Tour, under the Waste and Energy track.
Our mission is simple: recover reusable components, preserve their value, and make their second life verifiable.
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article