If I started learning AI Engineering in 2024, here’s what I would do.

The exact path I would choose.

In May 2023, I decided to go all-in and invest my time learning AI.

I took 6-months off (for fathers) and dedicated my time to learning about AI Engineering topics.

Fast forward 12 months, I have been working as a freelance AI Engineer, doing my first paid project.

I’m writing this article for aspiring AI Engineers. I’ll share my lessons, mistakes, and experience, because…

I want you to get where I am today (but much quicker).

By the end of this article, you’ll know:

what to learn
why to learn it
how to learn it faster
how to multiply the benefits of your learning

Disclaimer. 12 months ago, I didn’t start from scratch. I started exploring Machine Learning and Natural Language Processing in 2019. I’m also a Software Engineer. So, I had a decent handicap when I decided to transition into AI.

7 Steps to land your first AI Engineering job.

Let me share the exact steps I’d follow if I started learning AI Engineering from scratch.

Step 1: Learn basic Python.

AI Engineers must know how to code.

But to start, they need just some basics, like: ✔ Loops ✔ Variables ✔ Functions ✔ Data Types ✔ Basic Syntax ✔ Basic Operations ✔ Conditional Statements ✔ Usage of APIs and libraries ✔ Data Structures: ↳ Lists ↳ Tuples ↳ Dictionaries

It may sound like a lot, but it’s not.

Of course, the more proficient in Python you are, the better. But, your first goal is to learn the basics of building your first AI project (more on projects in step 4).

Luckily, the list I provided above is already enough to start.

I recommend these awesome Python resources:

Step 2️: Learn Large Language Models (LLMs).

LLMs are the “brains” of AI projects. AI Engineers use LLMs a lot! So, it’s important to understand how they work.

To start using LLMs, you don’t need to go deep. Today GPT-4, Claude 3, and Gemini are the most powerful LLMs. And even non-technical people use them with great success. It’s because they learned Prompt Engineering (more in step 3).

Again, to start applying LLMs in your projects, you don’t need to know exactly how they work.

Just understand these terms: ✔ tokens ✔ context window ✔ attention mechanism ✔ temperature (in LLMs) ✔ transformers architecture

They will give you the foundation to build your first AI projects. You will expand the foundations with every single project.

My favorite resources to learn about LLMs:

3. Learn the principles of Prompt Engineering.

AI Engineers know how to communicate well with LLMs.

So Prompt Engineering is a part of their job.

That being said, don’t spend too much time learning how to prompt. Don’t go too deep.

Just: ✔ learn how to be specific ✔ learn how to provide context ✔ learn how to be clear on your goals ✔ improve your prompts in iterations

I don’t believe Prompt Engineering is the skill of the future. The reason for that is simple: prompts are plain text. Writing great prompts already gets automated by LLMs and frameworks, such as DSPy.

That’s why I recommend NOT investing too much time here. But fundamentals are crucial.

Only when you need it, learn advanced prompting (resources are plenty).

You can learn prompt engineering here:

4. Build an AI project.

“Knowledge is best applied through execution.” Tiago Forte Building AI projects is the ONLY requirement to get hired as an AI Engineer.

The theory is not enough.

You can’t call yourself an AI Engineer unless you show you can actually create things using AI and code.

Luckily, you’re ready to build your first AI project (after completing the first 3 steps from this article).

Heck, if it’s your style, you can jump into the first project without going through the first 3 steps. You’ll learn them WHILE creating the project.

The first project should be simple.

We’ve got “only” 2 goals here:

To combine the first 3 steps.
To learn through practice (not just theory).

In your first project, you’ll use Python, LLMs, and prompts. It means you’ll combine everything from the first 3 steps. It’s simple thanks to OpenAI API or frameworks such as LangChain or LlamaIndex.

Let’s talk about learning through practice… “Never let the knowledge outpace execution.” Dan Koe

You don’t really learn unless you apply it.

I can’t stress it enough.

Look, I’ve wasted hundreds (if not thousands) of hours learning just theory. Today, I don’t remember 95% of that.

Around 3 years ago (at age 34), I decided to never learn anything without creating.

The results? ✔ I learn much faster. ✔ I document my learning. ✔ I remember almost everything. ✔ I gain a much deeper understanding. ✔ I build a portfolio of completed projects.

Note: I do 1 more thing that improves all these results (more in step 5).

So please, don’t make my mistakes.

Prioritize execution, not theory.

Going back to the project. My recommendation for the first one is to build the ChatGPT “clone”.

All you need to build the clone: ✔ Python ✔ OpenAI API ✔ Streamlit (for GUI) ✔ Around 40 lines of code

Sounds doable?

This project won’t be spectacular. But it will:

Put you in the “creator mode”.
Help you solidify knowledge from steps 1, 2, and 3.
Allow you to experiment with LLM parameters and prompts

If you need help with the project, you can use my article as your guide:

Create a ChatGPT Clone with Streamlit and OpenAI API: A Step-by-Step Guide

In 47 Lines of Code. From Scratch.

generativeai.pub

5. Learn in public.

“Show Your Work” by Austin Kleon is a 2-hour read.

Because of the book, I am writing this article right now.

It helped me understand that everybody has something valuable to share. Even as a complete beginner, you are ahead of most people.

Want a proof?

I just shared a link to the article I wrote ~6 weeks into my AI journey.

What to share?

✔ Code ✔ Lessons ✔ Mistakes ✔ Thoughts ✔ Favorite resources

Anything that you learn.

I know that going public is hard for most (it was scary for me too).

But it will be one of the best long-term investments you’ll ever make. The creator economy is booming. Creating content has almost no downsides (except for time investment).

Look what you’ll achieve: ✔ build trust ✔ inspire others ️✔ gain credibility ✔ build authenticity ✔ track your growth ✔ solidify knowledge ✔ grow your network ✔ learn deeper by teaching ✔ build your personal brand

And this list isn’t even complete!

I’ve done it, so I confirm the benefits.

So my tip for you is:

🌟 Share what you learn!

6. Repeat steps 4 & 5.

Keep building. Keep sharing.

How NOT to choose the next project?

This is a crucial question to ask before you jump into the next projects.

I made a huge mistake doing my projects because I had only a couple of days to build them. I published videos weekly, so I had to:

Find a project idea.
Select the technology.
Build the project (write code.
Record myself talking about the project.
Edit my recordings (hated it).
Prepare the description.
Upload the video.

Then, I wrote the article for Medium.

All within 7 days.

I was exploring and sharing rather than building real projects. Exploring itself is a good thing. But I wish I knew this “ugly truth” sooner:

Exploratory AI projects don’t get you hired!

How to choose the right project?

“Benefits, not features” is one of the main principles in sales.

Why do I bring up sales principles in the article about AI Engineering? Many people believe AI is overhyped.

AI news mentions cool AI features.

But businesses care only about results. Sadly, features rarely provide real benefits to businesses.

If you want your projects to stand out, focus on the benefits.

Here are the 3 main benefit categories:

saving time
saving money
making more money

(Hint: AI is great at saving time.)

Ensure the project adds value (provides real-world benefits).

go out of your comfort zone
ensure you’ll learn something new
improve your existing skills
solve real-problems

Each project should be challenging and force you to learn something new.

Here are some practical topics to learn: 🟢 RAG 🟢 LangChain 🟢 HuggingFace 🟢 Vector Databases 🟢 Vector Embedding 🟢 Open-source LLMs 🟢 image generators (DALL-e 3) 🟢 voice models (Whisper and TTS)

I’ve had 10+ job interviews for AI Engineering roles.

From my experience, companies always ask about RAG. So I’d use RAG in my 2nd or 3rd project.

The sooner, the better.

Also, it should improve your skills from steps 1, 2, 3. So you’ll get even more out of your projects when you:

expand your coding skills
learn more about LLMs
explore prompting

Build projects that solve actual problems.

7️. Find a mentor.

Not having a mentor has terribly slowed down my progress.

In my journey, I’ve done all the first 6 steps from the post (to some extent).

But I’ve never found a mentor…

I’ve never had a person that would: 🟢 help me go through hard times 🟢 tell me which skills to develop 🟢 help me with my career path 🟢 introduce me to his network 🟢 celebrate success with me 🟢 give me career guidance 🟢 give me feedback 🟢 encourage me 🟢 criticize me

Don’t make my mistake!

Find a mentor.

A mentor will save you months!

Note: If you’re looking for a mentor in the AI Engineering journey, I can help. DM me on LinkedIn (mention you come from Medium).

Conclusions.

It’s still extremely early to become an AI Engineer.

The role is only being defined. But the demand is quickly increasing.

After reading the article, you learned my path to becoming an AI Engineer in a record time.

If you have any questions, just write them in the comments! ✍️

🔔 My name is Kris. I help tech people become AI Engineers. On Medium, I share my experience and lessons that are valuable for aspiring AI Engineers. Make sure to follow. 🔔