Beyond the AI Score: Building ReLoop, a Circular Electronics Engine on AWS
Can AI help us decide whether an old device should be repaired, resold, donated, or recycled? Learn how we built ReLoop using Amazon Bedrock, AWS Lambda, API Gateway, Amazon S3, and DynamoDB—and why we separated AI-assisted assessment from deterministic decision-making to turn device images into actionable next-life recommendations.
An old laptop sits in a drawer. Its owner doesn't know whether repairing it is worth the cost, whether someone else could use it, or whether recycling is the only sensible option.
The device may still have value. The problem is knowing what to do next.
That question inspired ReLoop, an AI-powered circular electronics platform we built to help people make more informed decisions about unwanted devices.
Our central idea is simple: don't stop at identifying a device. Help people decide what its next life could be.
Building that experience introduced an interesting engineering challenge. We needed AI to help interpret device images, a consistent decision engine to evaluate possible pathways, and cloud infrastructure to connect the entire workflow.
Here's how we approached it using AWS.
1. From electronic waste to a decision problem
Electronic waste is not just a disposal problem. It is also a decision problem.
A device might be suitable for repair, useful to another person, valuable on the resale market, or ready for material recovery. The right choice depends on its condition, remaining utility, estimated costs, and the user's circumstances.
A generic AI-generated response would not be enough for our use case. We wanted a structured assessment that could compare multiple outcomes and explain a recommendation.
That led us to design ReLoop around six possible pathways:
- Repair: Fix the device and extend its useful life.
- Refurbish: Restore it for further use.
- Reuse: Find another practical purpose.
- Donate: Pass a usable device to someone who needs it.
- Resell: Explore recovering financial value.
- Recycle: Pursue appropriate end-of-life material recovery.
ReLoop analyzes an uploaded device image, produces a Next Life Score, and evaluates these pathways through its decision engine.
The experience is designed for students, households, and individuals who have old or unwanted electronics but need help understanding their options.
2. The architectural decision: AI is not the whole decision engine
One of the most important design choices in ReLoop is separating AI-assisted assessment from deterministic scoring.
AI is useful for interpreting an image and helping inform a device assessment. However, a recommendation involving money, device utility, or end-of-life decisions should not depend entirely on unconstrained model-generated text.
We therefore use two complementary components.
The AI-assisted assessment helps inform our understanding of the device.
ReLoop's deterministic decision engine applies structured logic to calculate the Next Life Score and evaluate the available pathways.
This separation gives the application a clearer division of responsibility. AI contributes to the assessment, while the decision engine produces a structured evaluation.
It also gives us a more maintainable foundation: assessment quality and decision logic can be examined and improved as distinct parts of the system.
A score alone is not proof that a recommendation is correct. Our broader goal is to make the reasoning, assumptions, and trade-offs behind that recommendation understandable to the user.
3. How AWS powers ReLoop
We use AWS managed services to connect the application interface, image handling, AI-assisted assessment, backend processing, and persistent storage.
Our core AWS stack includes Amazon API Gateway, AWS Lambda, Amazon S3, Amazon Bedrock with Amazon Nova 2 Lite, and Amazon DynamoDB. Amazon CloudWatch supports monitoring and troubleshooting.
ReLoop's architecture at a glance
The following diagram illustrates the major components and their roles in the assessment workflow.
At a high level, the frontend submits requests through API Gateway. Lambda coordinates backend processing, S3 provides image storage, and Amazon Bedrock supports AI-assisted assessment. ReLoop's deterministic decision engine calculates the score and evaluates the possible pathways, while DynamoDB stores assessment data.
CloudWatch supports monitoring of backend activity through the available logs and monitoring data.
The diagram represents the logical architecture. The precise sequence of service interactions depends on the operation being performed.
Amazon API Gateway: Connecting the application
Amazon API Gateway provides the entry point for requests from the ReLoop frontend to the backend.
It allows the application to communicate with backend processing through defined API endpoints, separating the user-facing experience from the underlying assessment workflow.
AWS Lambda: Running the backend
AWS Lambda runs backend processing for the assessment workflow.
This serverless approach lets us execute application logic without managing a continuously running application server for this layer. Lambda coordinates the relevant processing steps and connects the different parts of the workflow.
Amazon S3: Handling device images
Image-based assessment requires a way to store uploaded images.
Amazon Simple Storage Service (Amazon S3) provides object storage for device images used by ReLoop's assessment process.
Separating image storage from application logic creates a clear boundary between uploaded media and structured assessment results.
Amazon Bedrock and Amazon Nova 2 Lite: AI-assisted assessment
Amazon Bedrock provides the managed foundation-model access we use for AI-assisted image assessment with Amazon Nova 2 Lite.
This component helps inform the evaluation of an uploaded electronic device. Its output contributes to the wider assessment workflow rather than replacing the decision engine.
For us, the value of using AI is not simply generating a description of an image. It is making image-based information useful to a larger application that helps people evaluate their next steps.
Amazon DynamoDB: Persisting assessment data
Amazon DynamoDB stores scan results and assessment data produced by the workflow.
This provides persistent storage for structured results, separate from the image objects stored in S3.
Amazon CloudWatch: Monitoring and troubleshooting
Monitoring is essential when several services participate in a single workflow.
Amazon CloudWatch supports inspection and troubleshooting of backend activity through available monitoring data and logs. This helps us investigate backend behavior as we develop and maintain the application.
Together, these services provide the AWS foundation supporting ReLoop's assessment workflow.
4. Following an assessment through the system
The overall workflow connects several stages:
- A user uploads an electronic device image through the ReLoop application.
- The request enters the backend through Amazon API Gateway.
- AWS Lambda coordinates the relevant backend processing.
- Amazon S3 provides storage for the uploaded image.
- Amazon Bedrock with Amazon Nova 2 Lite supports AI-assisted assessment.
- ReLoop's deterministic decision engine calculates the Next Life Score and evaluates the six pathways.
- Assessment data is persisted in Amazon DynamoDB and made available to the application.
The key architectural principle is that the model, application logic, decision engine, and storage have distinct responsibilities.
That structure makes the system easier to reason about than treating the entire experience as a single AI prompt.
5. Making the result useful to a real person
A technical assessment is only valuable if a user can do something with it.
That is why ReLoop extends beyond image analysis and a numerical score.
The platform includes evidence breakdowns that help users understand an assessment, repair-versus-replacement comparisons that frame practical trade-offs, and INR-based estimates that are more relevant to users in India.
Personalized action plans help translate recommendations into potential next steps. Destination discovery helps users explore possible destinations for their devices, while a Decision Receipt provides a shareable record of the assessment and its recommendation.
These features address a critical distinction: identifying a possible next step is not the same as completing it.
For example, recommending donation does not mean a device has been donated. Estimating resale value does not guarantee a sale. A repair estimate does not guarantee that repairing the device is economically worthwhile.
We want ReLoop to communicate possibilities and assumptions clearly rather than present estimates as guaranteed outcomes.
6. Why transparency matters
Circular electronics involves several kinds of value.
A device can have financial value if it is resold, practical value if it remains useful, social value if it is donated, or material value when it reaches an appropriate recycling pathway.
These values are not interchangeable. Donation utility, resale proceeds, retained value, and material recovery should not be presented as if they were the same financial metric.
ReLoop's approach is to make estimates and assumptions explicit and help users compare pathways instead of presenting a single number as the complete answer.
This is particularly important when moving from a general AI assessment to a recommendation that could influence a real financial or environmental decision.
The score is a decision-support tool, not a guarantee. The recommendation is a starting point for action, not proof of a completed circular outcome.
We believe this distinction is essential to building trust in AI-assisted sustainability applications.
7. What building ReLoop taught us
Our experience reinforced three lessons.
First, a useful AI application needs more than a model call. The surrounding workflow—input handling, structured processing, storage, and presentation—is what turns model capabilities into a usable product.
Second, probabilistic AI and deterministic business logic serve different purposes. Separating them creates clearer responsibilities and gives developers more control over how assessments become recommendations.
Third, sustainability features need practical follow-through. Telling someone to recycle a device is less useful than helping them compare repair, reuse, donation, resale, and recycling before choosing a path.
AWS gave us the managed infrastructure to connect these parts into a working application, while the decision engine and user experience focus on the specific problem we set out to solve.
8. What's next for ReLoop?
We want to continue improving assessment quality, making pathway comparisons more informative, and helping users connect recommendations with real-world action.
As ReLoop develops, an important priority is keeping the relationship between an assessment and its outcome clear: what the system inferred, what it estimated, what the user chose, and what was actually completed.
That distinction will matter as much as the AI itself.
Every device deserves a better question
The question surrounding an old electronic device should not always be, "Where do I throw this away?"
It should be, "Could this be repaired? Could someone else use it? Is there value worth recovering? And if it has reached the end of its useful life, what is the responsible next step?"
ReLoop brings AI-assisted assessment, deterministic decision-making, and AWS cloud services together to help people explore those possibilities.
We are building toward a future in which the next step for an electronic device is a considered decision rather than an afterthought.
Don't just identify electronic waste. Discover what comes next.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article