How to become a freelance ML engineer in 2025?

While the number of ML job openings has exploded in the past years, the number of applicants to these jobs has exploded 10 times more.

Which means that landing an ML engineering job today is WAAAY harder than 5 years ago.

This makes you feel anxious, and can cause you a feeling of "I-am-not-good-enough-for-this".

This is called impostore syndrome.

And you know what?

Everyone has it. E-V-E-R-Y-O-N-E.

So, take a deep breath.
Stop for a second.
Make yourself a tea or a coffee, or whatever you like drinking.
And look at the problem from this angle…

 

What do companies posting these jobs try to solve?

Put yourself in their shoes.

Even for technical guys, like you and me, it has become extremely hard to follow the signal behind all the noise around AI.

Can you imagine how hard it is for non-ML guys, that lead companies, to understand and how can ML bring real business value for them?

How lost do you think these guys are?

These guys are in desperate need of technical people (like you and me) who can help them design and build ML software (with or without LLMs) that transforms their private raw data into smart operational decisions.

What does this mean for you?

You need to show you can frame a real world business problem as an ML problem. And then solve it, which means writing a Python service that can either

• process data (feature pipeline)

• train/fine-tune a model (training pipeline)

• serve the model's predictions (inference pipeline)

Learn to put it inside a Docker container, and deploy it to a compute platform, like AWS Lambda or Kubernetes.

This is what real world ML engineering is all about.

No more. No less.

And this is something you can only showcase if you build a PROFESSIONAL side project.

What does PROFESSIONAL mean?

The days when a cleaned up Jupyter notebook on your Github, gave you the job are GONE. This is not a sign you are good anymore.

You need to go further, and solve a specific business problem, by building and deploying an ML model (again, with or without LLMs).

This is what has value to companies.

Be brave, and

• Pick a real world problem that interests you

• Find a data API, ingest the data, and transform it into ML features

• Build a model (e.g. training an XGBoost model, building an LLM agent or fine tuning a base LLM),

• Build an API to serve the model predictions.

Clear thinking, and execution.
And of course a professional README file on your repo.

This is what the people hiring ML engineers needs and wants to see.

And this is something you CAN do.

And remember:

You don’t learn and then build.
You learn as you build.