I don’t just want to train a model in a Jupyter notebook and call it an ML project.
I want to understand how to take that model and turn it into something people can actually use.
Data → Model → API → Backend → Database → Deployment → Production
I want to understand every step.
And this is where I’ve started thinking about AI-assisted development.
Vibe-coding has become incredibly popular, and what I’d call “AI brain-rot” can become a form of cognitive offloading — where we stop solving problems ourselves because a tool can solve them for us.
And I don’t think cognitive offloading itself is necessarily bad.
The problem starts when we stop trying to understand what we’re building.
Because, in my perspective, something fundamental about engineering disappears when a project becomes:
Ctrl+C → Ctrl+V → It runs → Deploy.
But where is the joy?
Where is the satisfaction of figuring out why something works?
For me, software engineering is about taking something that exists in my head and turning it into something that exists in the real world — something useful, something people can interact with, something I can look at and say:
“I built this.”
And thinking about what ML can eventually do honestly gives me goosebumps.
I’ve always been a romanticist and a huge sci-fi movie fan.
TARS from Interstellar and JARVIS from Iron Man have probably influenced my interest in AI more than I realized.
Because they represent something fascinating to me:
Technology that doesn’t just execute commands, but understands context, assists its human, makes decisions, and interacts with the world around it.
Now imagine combining that kind of intelligence with modern AI, robotics, and increasingly capable autonomous systems.
Maybe one day we’ll build systems that feel much closer to those fictional assistants.
Maybe they’ll become AI companions, assistants, or something we haven’t even imagined yet.
I don’t know exactly what the future will look like.
But I know I want to understand how these systems are built.
I don’t want AI to think instead of me.
I want AI to make me a better engineer.
And I don’t want to merely use the AI revolution.
I want to understand it — and eventually build a part of it.
Top comments (8)
But those models counted on YOU!
You promised them one day they'll become like David Gandy or Kate Moss.
And you'll just leave them in the dark maze of the fashion industry, scared and alone?!
You were supposed to disrupt La Mode at the very heart of Île-de-France!
That is the job of a model trainer, Jonathan!
Vogue taught us bravery: to never run away from the runway, Sir!
HAHAHA 😭
I sincerely apologize to the models.
My entire career has been a misunderstanding here madame.
I thought I was deploying ML(machine learning?) models to production, but apparently I was supposed to be getting them onto the runway, but maybe I'll try my best anyways here?
I promise I’ll do better.
Data → Model → Runway → Vogue → Production 🫡
Good luck!
Don't forget to smile!
And if you have a con, turn it into a pro!
Zuckerberg is a counter-example.
Karate gi?! What is this home office Cobra Kai nonsense?
His images look like they were posted by a proud mother at a downtown dojo, clapping for a white belt, oblivious that it is like a free tie that the waiter gives you at a classy restaurant, if you show up in flip-flops and whatnot.
The Zuck must be himself!
Imagine a fragrance series:
Zuckerberg's Area 51, Abduct Your Senses
Training foundation weights has turned into a capital allocation game anyway. Frontier pre-training budgets are mostly consumed by distributed networking and hardware uptime rather than novel architecture.
The actual engineering complexity has moved downstream to the harness. Treating a model as an unreliable component inside a deterministic loop (managing latency budgets, state serialization, and failure boundaries) requires far more systems discipline than fitting a curve in a notebook.
Exactly right, that's what i call the engineering part is all about.
The model is just 1 part . The real challenge is how I can design a system/software where that model can be reliable and efficient. That is exactly why I have been diligently going slowly through the pathway of FastAPI-->PostgreSQL --> Machine Learning .
I want to understand what happens when that Model has to actually serve real-end users.
That exact downstream engineering is where I want to get better.
What about you ?
It is good to be idealistic; It has it's own benefits. But, for people like me who are graduates and actively-searching for jobs. Specialization in 1 is what we need . Plus , have tons and tons of things to learn even within the job itself. The bar is quite high i would say.
I understand that Copy-paste feeling . But , when the world demands pace and active-delivery .. vibe-coding becomes natural my friend.
Some comments may only be visible to logged-in visitors. Sign in to view all comments.