From Notebook to the Internet
Week 6 of my ML engineering journey, I pushed my first Dockerized ML model to the cloud.
Training a model feels cool.
Containerizing it feels powerful.
But deploying it?
That feels like magic.
This week, I watched my machine learning model leave my laptop and go live, running on Render, powered by Docker, reachable by anyone on the internet.
It wasn’t easy.
It wasn’t clean.
But it was real.
Last week, I wrapped my FastAPI model inside a Docker container.
This week, I decided to make it public.
The idea was simple:
If a model lives in my terminal, it only exists for me.
If it lives on Docker Hub, it can exist for anyone.
docker build -t nicelydone3/wallet-classifier:latest .
docker push nicelydone3/wallet-classifier:latest
That single command docker push changed everything.
My local project was now a versioned image on the internet.
A portable box of code, dependencies, and logic that could be pulled, run, or deployed anywhere.
It’s the closest thing to uploading your thoughts into the cloud.
Then came Render which was the final test.
I connected my repo, set up environment variables, exposed port 9696, and hit Deploy.
Five minutes later, logs started rolling.
CPU spikes, build output, success messages.
Then the moment:
“Service deployed successfully”
I copied the endpoint.
Opened Postman.
Sent a request.
And my model replied, live, hosted, breathing.
That single response hit harder than any accuracy scores I’ve ever seen.
When you train a model, you learn machine learning.
When you deploy it, you learn engineering.
Because now it’s not about accuracy.
It’s about uptime.
It’s about whether my API handles requests, survives restarts, and doesn’t collapse under bad input.
I stopped thinking like a data scientist.
Rather I started thinking like an engineer.
And suddenly, it all makes sense.
Pushing to Docker Hub and deploying on Render didn’t make me feel like I’d “finished a project”
It made me realize I’d started one.
Because now, my model doesn’t just exist in code, it exists in the world.
Accessible. Testable. Breakable.
And that’s the entire point.
Week six done.
The model’s alive.