
I Built Circuit Doctor: An AI That Looks at Your Circuit and Tells You What Went Wrong
What if you could simply take a photo of a circuit, upload it, and ask AI why it isn't working? That idea led me to build Circuit Doctor, a small AI-powered circuit troubleshooting assistant.
Why I started building Circuit Doctor
As an ECE student, I've spent quite a lot of time building small circuits on breadboards.
And one thing that almost always happens is this: the circuit doesn't work.
Then comes the usual process. Check the wires. Check the resistor. Check the LED polarity. Check the Arduino pin. Remove everything and connect it again. Sometimes the problem is something really small, but finding that small mistake takes much longer than expected.
That made me think about a simple question:
What if I could just take a picture of the circuit and ask an AI to help me figure out what might be wrong?
That question became the starting point for Circuit Doctor.
What is Circuit Doctor?
Circuit Doctor is an AI-powered circuit troubleshooting assistant that allows a user to upload a photo of a circuit and get an analysis of what is visible in the image.
The idea is simple.
Instead of manually explaining the entire circuit to an AI, the user can provide the image itself.
The application then tries to identify the visible components, understand the connections, point out possible wiring issues, and provide some troubleshooting suggestions.
For the first version, I kept the project small. I used a basic Arduino setup with an LED, resistor, breadboard and jumper wires as my test circuit.
The goal wasn't to build a perfect electronics expert.
The goal was to see whether I could turn a simple idea into a working AI application.
How it works
The basic workflow of Circuit Doctor is:
Circuit image → Upload → Backend → AI image analysis → Troubleshooting response
The user first uploads a photo of the circuit.
The application receives the image and sends it for analysis along with a prompt that tells the AI what to look for.
I wanted the response to be more useful than simply identifying components.
For example, if an image contains an LED and resistor, the application should also consider things such as LED polarity, whether the resistor appears to be connected properly, and whether there are suspicious or disconnected jumper wires.
The idea is to give the user possible things to check rather than just describing what is present in the photograph.
Using AI for something practical
I've experimented with AI through different projects and tutorials before, but this project was different because I wanted the AI to actually solve a problem.
The interesting part was not simply sending a question to a model and getting an answer.
The interesting part was figuring out how to take an image from a user, pass it through an application, give the model the right instructions, and then return the result in a way that makes sense to someone who is trying to debug a physical circuit.
That made me think more carefully about prompts as well.
A vague prompt gives a vague answer.
For Circuit Doctor, I wanted the AI to focus on components, connections, possible faults and practical troubleshooting steps.
Where AWS fits into the project
While building Circuit Doctor, I also wanted to understand how a project like this could move beyond just running on my laptop.
That's where AWS became an important part of the project.
The architecture I was working toward uses cloud services for handling the application workflow, image processing and storage, while the AI model handles the visual analysis.
The bigger idea is to have a workflow where a user can upload an image and the application can process it without depending on the user having a complicated local setup.
For me, this was one of the most useful parts of the project because I started thinking about the difference between building something that works locally and building something that could actually be used by other people.
Building it was not as smooth as I expected
The project definitely had its share of problems.
Getting the application running involved dealing with environment variables, API keys, packages, backend requests and image handling.
There were also moments where something looked like it should work, but simply didn't.
During my final testing, I eventually ran into a Gemini API quota error.
At first, the application only showed a generic analysis failure. After checking the technical error, I found that the request was actually reaching the Gemini API. The problem was that I had exceeded the available free-tier request quota for the project.
It was frustrating because it happened while I was trying to get the final demo ready, but it was also a useful lesson.
An application can have perfectly working frontend code and still fail because of an external service limit.
That is something you don't really understand until you build and test an application yourself.
What I learned from building it
The biggest thing I learned from Circuit Doctor is that building a project is very different from just learning individual technologies.
Before starting, it is easy to think of a project as:
"Frontend + AI = application."
In reality, there are many small pieces that have to work together.
I had to think about the user interface, image uploads, backend communication, API authentication, environment variables, prompts, error handling and cloud infrastructure.
I also learned that the first version doesn't need to solve everything.
It is better to build a small working idea, test it, find the problems and improve it than to spend weeks designing a huge system without actually building anything.
What I want to improve next
Circuit Doctor is still an MVP, so there is a lot that can be improved.
One of the biggest improvements would be making the AI better at understanding the actual circuit rather than just recognizing objects in an image.
I would also like to support more components such as sensors, motors, transistors and different types of ICs.
Another important improvement would be giving the AI more context.
For example, a user could tell Circuit Doctor what the circuit is supposed to do and what is currently happening. The AI could then combine that information with the image to provide more useful troubleshooting steps.
I would also like to improve the cloud architecture and deploy the application properly so that other students can use it without needing to run everything locally.
Why this project matters to me
I didn't start Circuit Doctor with the idea of building a huge product.
I started with a problem I had personally experienced while working with circuits.
As an ECE student, I am learning electronics, but at the same time I am also trying to learn software development, AI and cloud technologies.
Circuit Doctor gave me a reason to bring all of these areas together in one project.
It also showed me something I think is important when learning technology:
You understand things much better when you actually try to build something with them.
Reading about APIs is one thing.
Debugging an API at midnight because your project isn't working is a completely different experience.
What's next for Circuit Doctor?
The current version is just the beginning.
My next goal is to make the analysis more reliable, support more types of circuits, improve the troubleshooting suggestions and move the application toward a proper cloud deployment.
Eventually, I'd like Circuit Doctor to become something that an electronics student could actually keep open while working on a project.
Take a picture.
Get a second opinion.
Check the suggested connections.
Fix the circuit.
That is the simple idea behind Circuit Doctor.
And this project started with a question that I've probably asked myself more times than I'd like to admit:
"Why isn't this circuit working?"
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article