How to learn Machine Learning

Let me show you the strategy I use to learn new things in Machine Learning, and stay relevant in the field.

The problem → The need to *constantly* learn ML 🔁

Machine Learning is an exciting field to work on, but it is also complex and fast-evolving. It does not matter how much you know today, your skills and expertise will become obsolete faster than you think.

This might seem discouraging at first. But, for me, this has turn out to be the main reason I work in ML.

The excitement and joy I experience when I learn a new thing, for example

• a new programming language, like Rust

• a new library, like Polars, or

• a completely new paradigm for building ML apps, like Large Language Models

is what makes ML special for me. And I am damn sure you feel the same.

However, learning new things in ML these days can be overwhelming. At the end of the day, there are so many online courses, blog posts, books and AI influencers to follow, that the question is

What is the best way to learn ML, and grow in this field ?

Let me share with you my strategy ↓

The solution → Project based learning 🛠️

I frame my learning always as a project. That is, I start by asking myself:

What project do I want to build with this new thing I want to learn?

When you start from the end goal you keep focus, avoid distractions, and learn tons of things pretty fast.

Example

Let’s say you wanna get up to speed with Large Language Models (LLMs). A good first project would be to build a REST API for sentiment analysis.

When you start your learning path with a clear goal in mind, you put your brain in a problem-solving mode. And you start asking yourself the right questions:

• Where can I find a dataset to train and evaluate my model?

• What is the right model for this task?

• How do I wrap and deploy my model as a REST API?

So you go to the Internet and start looking for answers. At first you feel a bit overwhelmed. But eventually you start connecting the dots, and finding answers to your questions. For example,

• Dataset? → This Amazon Review dataset can be a good starting point .

• Model? → The Hugging Face transformers library provides great models for sentiment analysis, which you can use out-of-the box or even fine-tune with your own data to improve their accuracy for your task.

• Deployment? → AWS Lambda is a popular serverless deployment platform, why not try it?

These are the building blocks of your solution, that you need to put together to be a champion.

But it's been no bed of roses 🛏️🌹

Project-based learning is not a happy path, because having the ingredients is not enough to build your solution. You need to get your hands dirty, and start writing the code that cooks these ingredients into your masterpiece.

This phase is full of ups and downs, little victories and frustrations. And this is when REAL LEARNING takes place.

Only when you struggle and implement solutions yourself, you learn. As simple as that.

We are the champions, my friends 🏆

When you based your learning on building projects you get 2 things:

• Immense joy when you manage to overcome obstacles and complete the project.

• An incredible portfolio, that you can showcase in job interviews, that will help you stand out from the crowd.

And before you leave…

Wanna learn to build real-world ML apps with me?

This week 100+ brave students and myself finished building a real-time ML system to predict short-term crypto prices.

It has been an exciting 5-week trip (25+ hours of live coding lessions), in which we have used a project based approach to learn how to

• Design any ML system, using Feature-Training-Inference pipelines and a mesh of microservices.

• Write Professional / well-packaged Python Code using Poetry and Docker.

• Build real-time data processing pipelines in Python.

• Integrate a Feature Store (Hopsworks), a Message Bus (Redpanda), a Model Registry (Comet ML) and a Compute Platform (Quix Cloud) to deploy the whole system.

Wanna read all the details? → Click here to learn more

And the best part?

Join today and have access to

→ all future sessions of the program, and

→ 25+ hours of live coding sessions and the full source code implementation of the system we’ve built in the past 5 weeks.

Looking forward to having you in the next cohort of the course
↓↓↓

Have a great weekend,

And talk to you next week

Pau