🗺️ Your Roadmap to Become an AI Engineer in 2024

Introduction

AI engineers are all the rage right now. Seriously, everyone’s talking about them. Why? Because AI is driving everything.

If you’ve ever thought about jumping ship from your software engineer gig and diving into the world of AI, you’re in the right place.

This guide is for anyone ready to make that transition—from coding a random blog app to building the future with AI. You’ll learn the key skills you need, the tools to get your hands on, and how to actually develop these skills without feeling like your brain is going to melt. Stick around; this is going to be fun (and possibly the best decision you’ll ever make, right next to choosing your favorite takeout).

I. Who is This Roadmap For?

So, who exactly is this AI Engineer roadmap for? Well, if you’re a programmer, SDE, data analyst, or data scientist who’s tired of just pushing code and want to really get into the world of AI (without losing your mind), this is for you.

Now, let’s get real for a second—you don’t need to be a genius, but here’s what you will need:

• Intermediate knowledge of:

  • Python or JavaScript

• Experience building:

  • 2-3 moderately complex applications (think blog projects or web apps)

• Comfort with:

  • Reading documentation (without your brain exploding)

  • Using IDEs like VS Code (no tears allowed)

  • Git and GitHub (but don’t worry, you can pick this up along the way)

If you’ve done all of the above without flipping your desk, you’re so ready for this roadmap. This is your chance to move from being a regular old software engineer to a certified AI Engineer.

Ready? Let’s keep rolling.

II. Breaking Down the Roadmap

Let’s talk about the journey to becoming an AI Engineer—because it’s a journey, not a sprint (and thank goodness for that, right?). This roadmap isn’t just about ticking off boxes; it’s about real progression. We’re starting with baby steps and moving toward mastering the AI world. It’s broken into three stages: Beginner, Intermediate, and Advanced—and yes, you will get smarter along the way.

1. Beginner (≤ 1 month)

This is where we start small. You’re getting comfortable with the basics, and honestly, this part is more about figuring out how to talk to AI without sounding like a robot yourself.

2. Intermediate (~2 months)

Now, we’re moving up. This stage is where you start feeling like you might actually know what you’re doing (emphasis on might). You’ll be tackling deeper concepts, and yes, things start getting a little more serious.

• Learn about Retrieval-Augmented Generation (RAG)—in simple terms, this lets your app understand the context instead of just spitting out random facts.

• Vector databases will become your new best friend (or worst enemy, depending on your patience).

  • This is where you start building AI that can hold an intelligent conversation.

• You’ll be building agents—little automated helpers that can handle tasks on their own.

  • Think of it as parenting, but less messy.

3. Advanced (~3 months)

Okay, now you’re not just in the AI world—you’re running the show. This stage is where you start building things that actually make sense (and might even impress someone other than your mom).

• Focus on deployment and optimization—because what’s the point of building something if it doesn’t actually work in the real world?

• Master LLMOps—AI’s version of DevOps, to ensure everything runs smoothly from start to finish.

  • You’ll also be fine-tuning pre-trained models to make them super specific to industries like medicine, finance, or even cat memes.

  • Yes, fine-tuning cat memes is a real job.

The progression from Beginner to Advanced isn’t about being perfect—it’s about learning as you go, figuring things out, and probably Googling a lot more than you’d like to admit. But hey, by the end of this, you’ll be a full-fledged AI Engineer, and that’s something worth celebrating!

III. Beginner-Level Skills

Alright, so you want to be an AI Engineer? Let’s start with the basics—because every expert was once a beginner (or at least that’s what I tell myself when I’m stuck at 2 AM). This is where you build your foundation.

1. Basics of LLMs

You’ve probably heard of ChatGPT, right? At this stage, you’ll need to know how it works—not the deep, scary stuff yet, but enough to get by. Think of it like understanding how your phone works without actually knowing how to code it.

2. Prompt Engineering

Talking to an AI is like talking to your dog—sometimes, it gives you what you want, and sometimes, it stares blankly. That’s where prompt engineering comes in. You’ll learn how to phrase things just right to get the answers you need (and avoid weird responses that make you go, “Huh?”).

3. Consuming APIs and Working with JSON

Ever connected your fridge to your phone and thought, “This is so cool”? No? Well, it’s kind of like that. APIs help you grab data from different places, and JSON is the format it usually comes in. So, you’ll learn how to fetch data and make it work for you—without the stress of restocking milk.

4. Calling Closed and Open-Source LLM Models

Think of closed-source models like your family’s secret recipe, and open-source ones as the recipe your friend gave you. Both are useful, but you’ll learn how to get them working for your app. Whether it’s ChatGPT or something else, you’ll be making calls and pulling data like a pro.

• Deeplearning.AI course on Open Source Models with Hugging Face

• Open Source LLMs can be accessed via Hugging Face Hub and you can play with a few of them in Hugging Face Spaces

• OpenRouter Docs

• Run LLMs on your local machine using LM Studio

5. Managing Context and Building Sequences

AI can have the memory of a goldfish if you’re not careful. That’s where managing context comes in. Using tools like Langchain, you’ll learn how to keep the conversation flowing so your AI doesn’t ask, “What were we talking about?” every five minutes. You’re basically becoming the AI’s memory coach—without all the motivational speeches.

6. Basic App Development

Now you’re cooking! With tools like Gradio or Streamlit, you’ll start building your first apps. They’re not going to change the world (yet), but they’ll make you feel like a wizard who can make things happen. Think of it as building your first LEGO house—small, but satisfying.

7. Deploying Apps

Built something cool? Now it’s time to put it out there for the world to see! You’ll learn how to deploy your app on platforms like HuggingFace Space or Streamlit Cloud. It’s like putting your project on display and saying, “Look, I made this!”—minus the glitter and poster board.

8. Multimodal Generation

Let’s add some flair to your skills. You’ll start dabbling in multimodal generation, where you combine code, images, and audio—all at once. Using HuggingFace libraries, you’ll orchestrate it like a conductor leading an AI symphony. It’s cool, it’s complex, and yeah—it’ll impress your friends (or at least your dog).

• This is covered in parts in the Open Source Models with HF course on DeepLearning.AI

• Building Generative AI Applications with Gradio

• Code Gen: Check out these resources on code generation – gpt-engineer, Tabby, gpt-migrate to migrate your codebase from one framework to another or one language to another.

• Audio Gen: text to speech by openAI, resemble.ai, elevenlabs API

• Image Gen: Image generation by Open AI, creating images using Stable Diffusion API.

By the end of this stage, you’re not just a beginner—you’re a beginner AI Engineer who can build things, troubleshoot, and show off your creations. And honestly, that’s where everyone starts. So keep going, keep learning, and when in doubt, grab another cup of coffee and keep Googling.

IV. Intermediate-Level Projects and Skills

So, you’ve made it through the basics, and now you’re ready to step up your game. Welcome to the intermediate level, where things get a little more complex—but also a lot more fun. You’re not just “playing around” with AI anymore—you’re building things that start to feel real. Let’s break it down.

1. Vector Databases

Okay, so vector databases might sound like something out of a sci-fi movie, but trust me, they’re just fancy tools to store and retrieve data. As an AI Engineer, you’ll need to understand how to use these for AI applications. They’re like your trusty sidekick, keeping all the information in check while your AI works its magic.

• Text chunking and splitting by LangChain

• Course on vector databases: Learn what are embeddings and how to store them. Build applications.

• Another course on building applications with vector databases using Pinecone

• Learn to compute sentence, text, and image embeddings using Framework like SentenceTransformers.

• Check out top embedding models here.

2. Building RAG Systems

Next up, Retrieval-Augmented Generation (RAG) systems. Imagine having a knowledge-based chat system that doesn’t just throw random facts at you but actually understands context. Yeah, that’s RAG. You’ll be developing these systems to make your AI useful (because, let’s face it, no one wants to chat with an AI that’s clueless).

• RAG applications are all about building connections between tools, databases, context lengths, embeddings, memories, etc. You need frameworks like LangChain, LlamaIndex, FastRAG to build these.

• LangChain’s RAG from Scratch playlist on YouTube is pretty detailed and amazing.

3. Advanced RAG Pipelines

Now here’s where things get exciting (or stressful, depending on how much coffee you’ve had). You’ll be building advanced RAG pipelines, which include:

• Sub-question query engines: This means your AI can break down a complex question and fetch the right answers. It’s like giving your AI a few more brain cells to work with.

• Check out Jerrry Liu’s course on Building and evaluating Advanced RAG Application on DeepLearning.AI for best practices and improving your RAG pipeline’s performance.

• Cheat Sheet and some recipes for building Advanced RAG

• A comprehensive guide on building RAG-based LLM application by AnyScale

4. Multi-Agent Systems

Remember when you thought one agent was cool? Well, now you get to work with multi-agent systems. It’s like assembling a team of AIs, each handling different tasks. Think of it as the Avengers, but for AI—minus the capes, plus more data.

• Quickstart guide by LangChain to build agents to have a sequence of actions taken to do a job or multiple jobs.

• Course on Functions Tools and Agents with LangChain by Harrison Chase on DeepLearning.AI

5. Automating Workflows

You’ll start automating workflows with tools like Autogen and Crew AI. Imagine setting up a system where you don’t have to lift a finger once everything is in place. It’s like having your own little army of bots doing your work—just don’t let the power go to your head.

6. Evaluating RAGs

Building RAG systems is great, but you’ve got to know how well they’re working. Enter the RAGAs framework. This is where you evaluate whether your AI is performing as expected. And if it’s not? Well, time to tweak a few things (and maybe have a meltdown before fixing it).

• Hugging Face Cookbook on How to evaluate RAG system.

• Evaluating all components of your RAG pipeline.

• RAGAS framework to evaluate RAG pipelines.

7. Managing Databases and Deployment

Finally, you’ll be managing databases, retrievals, and deploying applications like a pro. You’ll also handle versioning and monitoring your models. It’s like being a project manager for AI—you’ve got to keep everything running smoothly, or else it’s back to the drawing board.

• Local deployment: Running open source LLMs on local machines (LM Studio, Ollama, oobabooga, kobold.cpp, etc.)

• Building POCs and demo applications using frameworks like Gradio and Streamlit.

• Deploying LLMs at scale on cloud technologies like vLLM and SkyPilot.

• Deploying LangChain applications (runnables and chains) as a REST API.

This stage might feel like a lot—and it is—but once you get through it, you’ll be an AI Engineer who can handle projects that actually make a difference. It’s not all rainbows and butterflies, but the best things rarely are. And just like that time you cried for three hours in a coffee shop and still managed to pull off three killer posts, you’ve got this. Keep going!

V. Advanced-Level Projects and Skills

Alright, you’ve made it to the big leagues. At this point, you’re not just an AI Engineer—you’re becoming the AI engineer everyone else turns to when things get complicated. These advanced projects aren’t for the faint of heart, but you’re not here because you like taking the easy way out, right? Let’s get into the nitty-gritty of what makes this level different.

1. Fine-Tuning Pre-Trained LLMs for Specific Domains

Imagine being able to teach an AI model the ins and outs of a specific field—whether it’s medical, legal, or financial. That’s what fine-tuning pre-trained LLMs is all about. You’re not just using generic models anymore; you’re making them experts. It’s like training a generalist into a specialist, except you don’t have to deal with a temperamental grad student.

2. Curating and Engineering Datasets (ETL Pipelines)

Welcome to the world of ETL pipelines (Extract, Transform, Load). You’ll be curating datasets and making sure they’re just right for your AI to learn from. Think of it as creating the perfect playlist, but instead of music, it’s data that’ll make your models better and smarter. Yes, you’re essentially the DJ of data now.

• DeepLearning.AI course on finetuning LLMs.

• A Beginner’s Guide to LLM Fine-Tuning is a detailed guide on finetuning LLMs.

• A very detailed and simplified read on how to fine-tune LLMs with Hugging Face by Philipp Schmid.

• 4-part blog series by Anyscale is a comprehensive guide on fine tuning and serving LLMs.

• Fine-Tune Your Own Llama 2 Model in a Colab Notebook.

• LLM Datasets for finetuning.

3. Evaluating and Benchmarking Model Performance

How do you know your model is actually good? Well, you evaluate and benchmark it. You’ll run tests, compare results, and figure out if your AI is top-tier or if it’s still hanging out in the “needs improvement” zone. It’s kind of like grading papers, except the papers are AI models, and you really don’t want to see a D.

4. LLMOps: Building Complete Pipelines

Here’s where things get serious: LLMOps is like DevOps for AI. You’ll be building full pipelines, including model registry, observability, and testing. It’s your job to make sure the entire AI workflow is smooth, from data collection to deployment. And no, you won’t get a break if things break down (pun intended). You’ve got this!

• Deeplearning.AI Course on LLMOPs is a good starting place for advanced practitioners.

• GPU Inference optimization techniques like FlashAttention and FlashAttention-2.

• LLMOps guide by DataBricks.

• Efficiently serving LLMs course on DeepLearning.AI.

5. Multi-Modal Applications

Ever wanted to combine text, images, and maybe even some audio into one cohesive AI app? Now you can. You’ll be building multi-modal applications—think hybrid semantic search that can process text and images at the same time. You’re officially stepping into the world of super fancy AI magic. Time to impress everyone with your wizard-like abilities.

• Building hybrid search apps with vector databases like Pinecone.

• Cookbook for multimodal RAG pipelines.

6. Creating SDKs and Custom Solutions for Developers

Here’s where you get to be the hero for other developers: creating SDKs and custom solutions. These will let other people build on what you’ve created, which is basically your way of saying, “Hey, I built something awesome, and now you can too.” It’s like being the cool kid who shares their toys on the playground.

• How to evaluate LLMs with Hugging Face Lighteval.

• Course on Automated Testing for LLMOps: Learn to test and evaluate LLM application using an evaluation LLM.

7. Securing AI Applications Against Vulnerabilities

AI isn’t all fun and games—there are security risks, too. At this stage, you’ll learn how to secure AI applications against things like prompt hacking and other vulnerabilities. Think of it like building a fortress around your AI to make sure no one sneaks in and messes with your model. You’re basically becoming an AI bodyguard.

• Red Teaming LLM Applications – learn to identify and evaluate vulnerabilities in LLM apps.

• Planning red teaming for large language models (LLMs) and their applications.

• A detailed list of resources on LLM security highlighting all potential risks and vulnerabilities in AI applications.

At this level, you’re not just working on AI—you’re shaping the future of it. Yes, it’s hard work. Yes, you’ll probably cry in a coffee shop again (I’ve been there), but at the end of the day, you’ll be proud of what you’ve built. You’re an AI Engineer through and through, and honestly? That’s something to celebrate. Keep going—you’ve got this!

VI. How to Develop These Skills

So, how do you go from aspiring AI Engineer to pro? Simple—build. Let’s break it down:

• Learn by Doing: Reading alone won’t cut it. You need to build real projects. Whether it’s a chatbot or an AI model, get hands-on. Think of it like learning to cook—you’ve got to get messy to get better.

• Hands-On Experience: It’s the key. Start with small projects and gradually level up. You’ll learn more by doing than by reading endless documentation.

• Check the Project Repository: Stuck for ideas? Explore the repository full of project ideas, from beginner to advanced. It’s a treasure trove for anyone looking to sharpen their skills.

• Keep Going: There will be mistakes (probably a lot), but keep building. Every project gets you closer to being an AI Engineer. You’ve got this!

Stay hands-on, keep learning, and remember—the more you build, the better you get.

Conclusion

Becoming an AI Engineer is no easy ride—expect to feel like your brain might melt at least once. But the truth is, every single moment of confusion, frustration, and even late-night Googling is part of the process. You’re not just learning to build AI; you’re building the skills that will make you a legit AI Engineer. And yeah, there might be days when you feel like throwing your computer out the window, but stick with it. Cry if you need to—just make sure you get back to work.

Remember: even when things feel impossible, you’ve got this. Because, let’s be real, who else is going to fine-tune AI for cat memes? Keep going—you’re closer than you think.

If you are interested in other topics and how AI is transforming different aspects of our lives, or even in making money using AI with more detailed, step-by-step guidance, you can find our other articles here:

• Automate Your Video Creation with AI: Full Code Inside for Fast and Easy Results!*

• AI vs Solo Developers: Who’s Really Building the Future?*

• FLUX AI: The Game-Changing Art AI Generator That’s Shaking Up the Industry

• The Secret to Earning $8K/Week with Canva and Free AI Tools!

• Create Your Own AI-Generated Short Film: Simple Steps to Bring Your Ideas to Life*

*indicates a premium content, if any