Here at ZTM, we’re focused on helping you learn the most in demand tech skills, so you can get hired or promoted into more senior roles.
Now obviously, AI is the most in demand area right now, but it’s kind of exploded with so many different areas of focus, which means it’s hard to know what to learn or why. Case in point, we have 50+ AI focused courses of our own!
So I figured I would help you out.
In this guide I’m going to break down each of our AI courses and what they’re for, so you get a better idea of what to learn next and where to upskill for AI in your career.
Let’s get into it...
Learn how to work with AI
You can open ChatGPT, type a question, and get an answer in seconds. So why would you need a course to teach you how to do that?
Because getting an answer is easy. But getting one you can actually use, and knowing when you shouldn’t trust it, takes a bit more understanding.
Tools like ChatGPT and Claude are built around large language models, or LLMs. These models have learned patterns from enormous amounts of data, which lets them generate useful responses to your questions and instructions. But they don’t know your job, your goal, or what a good answer would look like unless you give them that information. They can also produce an answer that sounds certain even when it’s wrong.
That’s why these first courses cover both sides of working with AI:
Understanding what the tool is doing
and learning how to guide and check its output
Introduction to Prompt Engineering
A prompt is the instruction or question you give an AI tool. If you’ve been typing a quick request and hoping for the best, Introduction to Prompt Engineering shows you what changes when you give the model a clearer goal and more useful information:
You’ll try tools such as ChatGPT and Claude, explore what they can do, and practice by getting AI to help you build a simple game. For this particular course you don’t need to know how to code. The point is to start noticing how your instructions affect what the AI gives back, rather than treating every response as a bit of magic.
How LLMs Work
Once you’ve used one of these tools, you’ll probably have questions like:
Why can it explain a difficult idea so well, then get a simple fact wrong?
Why might it answer the same question differently the next time?
How LLMs Work takes you underneath the chat interface and explains the ideas behind the model:
It helps you judge which tasks an LLM is suited to, recognize where its answer might be unreliable, and ask better questions when something seems off.
You don’t need to become an AI engineer to benefit from that knowledge!
Prompt Engineering Fundamentals
Understanding the model is helpful, but you still need to put that understanding to work.
For example
Say you ask AI to write a reply to a customer.
If all you told it was “Write a polite reply”, that leaves it to guess the situation, your company’s policy, and what the customer needs to hear. But if you give it those details, an example of the tone you use, and the outcome you want, and you’ve given it a much better chance of helping.
Prompt Engineering Fundamentals teaches you how to do that deliberately:
This course covers how to give the model context, show examples, structure instructions, and how to ask for an answer in a useful format, so you get the results you want. This a key a prompting skill to have for whenever you’re writing, researching, analyzing information, or eventually building an AI-powered tool.
Prompt Engineering Bootcamp
If you’d rather take a more complete route through the subject, the Prompt Engineering Bootcamp starts at the beginning and goes further into LLMs, prompting techniques, different models, and hands-on projects:
It overlaps with the shorter courses above, so you don’t need to finish all three before starting it. You can think of the shorter courses as ways to explore a specific question, while the bootcamp is for someone who wants to spend more time practicing how to work with LLMs across different situations.
I highly recommend this course regardless of your role, interests, or where you choose to specialize.
Use AI in the work you already do
Getting a useful answer in a chat is one thing, but getting it to help with the work that keeps showing up on your desk is even better. That's why these next courses take prompting out of the abstract and put it into familiar tasks.
ChatGPT for Excel
Say you’re looking at a spreadsheet and need a formula that checks several conditions. You could search for examples, piece a formula together, and hope you’ve understood it correctly. Or you could explain the result you need to ChatGPT, show it how your data is arranged, and ask it to help you build and understand the formula.
That’s the idea behind ChatGPT for Excel:
It shows you how to use AI to tackle spreadsheet problems, work with files, and break larger tasks into steps.
Build AI Agents with n8n
But what if the task involves several tools? Perhaps information arrives in an email, needs to be checked, added to a document, and sent to someone for review. A chat response can help with one part, but it won’t connect all those steps by itself.
n8n is a visual tool for connecting apps and arranging those steps into a workflow. Its triggers decide when the workflow starts, and its nodes carry out the individual actions. You can add an AI model where it’s useful, such as interpreting the incoming information or drafting a response.
The Build AI Agents with n8n course teaches you to put those pieces together without writing code:
Its a great course if you’re starting to spot repetitive processes in your work and want to do more than ask an AI tool for help each time. You’ll still need to check what the workflow does, especially before letting it take actions automatically, but learning this can make your life a lot easier.
Use AI to help you build software
So far, we’ve looked at using AI to help with general office tasks but what if the work you do is writing software?
Well, good news because AI can help you with that also - just as long as you learn how to use it correctly...
Vibe Coding Bootcamp
If you’re interested in building something but haven’t written much code, our Vibe Coding Bootcamp lets you start with an idea and use AI tools to help turn it into a working project. You’ll explore tools such as Cursor and GitHub Copilot while building websites and apps:
One of the big appeals with vibe coding is the fact that you can get something on screen quickly. But that also makes it tempting to accept code you don’t understand simply because the page looks right.
That's why the course puts planning, testing, and improving the result alongside generating it. It’s a good fit if you want to build projects and learn how to direct AI through the process, whether you’re new to coding or already a developer.
AI Coding with GitHub Copilot
What if you already write code and want AI help inside the editor you use every day? Well, GitHub Copilot can suggest code as you work, and its chat features can help explain code or draft documentation.
AI Coding with GitHub Copilot is a short course focused on using that tool in editors such as VS Code and PyCharm.
You can start small and ask it to explain a function, compare that explanation with what the code actually does, and then try it on a change you’re making yourself.
AI Coding with Jupyter AI
If you work with Python for data analysis or machine learning, you might use a Jupyter notebook, where you can run small sections of code and see their results alongside notes and charts.
Jupyter AI adds AI assistance to that environment, and the AI Coding with Jupyter AI course shows you how to set it up and use it while learning or writing Python.
Understand AI-Assisted Software Development
A coding assistant becomes more challenging to use when the project gets bigger.
Why?
Well, a change to one file can affect another. Generated code might pass a quick visual check while introducing a bug, an unsafe dependency, or something your team will struggle to maintain.
Our Understand AI-Assisted Software Development course focuses on the judgment and workflow behind using AI on real software projects:
It covers planning a feature, giving the tool the relevant project context, working through changes in stages, and reviewing and testing what it produces. This is particularly useful if you already develop software and want AI to help with substantial work rather than isolated code suggestions.
Claude Code Bootcamp
If you want to go deep on one tool for coding, our Claude Code Bootcamp focuses on using Claude Code to plan, build, test, and deploy web apps:
It starts with turning an idea into clear requirements, then shows you how to give Claude the project context and standards it needs as the build progresses. More specifically, it covers how to direct a project, not just asking for a piece of code. You'll learn how to inspect what was built, test it against what you asked for, and keep improving it.
Build applications that use AI
There’s a big difference between using AI to help you build an app and building an app that uses AI when someone opens it.
For example
Think of a note-taking app that turns a meeting recording into a summary. The developer doesn’t need to train a speech recognition model or a language model from scratch. Their app can send information to an existing model and use the response to give the customer a useful feature.
And this group of courses gives you several ways to learn how to do this.
Building AI Apps with the OpenAI API
If you want to see how an app talks to an AI model, our Building AI Apps with the OpenAI API course is a great starting point. It shows you how to send requests for text, images, speech recognition, and spoken audio, then use the results in an application:
You’ll build a meal-planning app as you learn. The user gives your app their preferences, then your app sends the relevant request to the AI service, and then it presents the result.
Handy right?
Once you understand that process, you can start thinking about AI features for your own projects.
Building AI Apps with the Gemini API
Our Building AI Apps with the Gemini API course teaches you the same broad skill as the previous course, but this time using Google’s models. The projects includes asking questions about an image and organizing photos based on what’s in them:
Why might you choose this course? Perhaps the feature you want to build involves images as well as text, or you want experience with another model provider.
However, you don’t have to take both API courses before you can build anything. Pick one, make something small, and come back to the other when you have a reason to compare them.
Developing LLM Apps with LangChain
Once an app does more than send one request and display one answer, you need to manage the steps between them.
For example
It might need to find a relevant document, pass part of it to a model, and return an answer to the user, and LangChain is a framework developers can use to connect those parts.
In our Developing LLM Apps with LangChain course, you build a question-answering application that uses external documents.
It’s a useful next course when you understand the basic API request and want to see how a fuller AI application fits together.
Developing LLM App Frontends with Streamlit
Your application also needs somewhere for a person to use it.
For example
A working Python script may prove your idea, but most people aren’t going to open a notebook and run code just to ask it a question. Well the good news is, our Developing LLM App Frontends with Streamlit course shows you how to create front-end intefaces.
In this course you'll give a question-answering app a front end, so someone can type a question and see the response:
Learn this course content for when you’ve got an AI feature working and want to make it easier to demonstrate or use.
AI Engineering: Building AI Applications
If your goal is to develop more complete AI applications, our AI Engineering: Building AI Applications course goes further into working with LLM APIs and tools such as LangChain, LangSmith, and LangGraph:
This is useful to learn for when you’re dealing with several connected steps, has more complex behavior, and the need to understand what the application did when something goes wrong.
You can start with a small API project to learn the basic exchange, then use this course when you’re ready to build a larger system.
Connect an AI app to information it needs
Suppose you build an assistant to answer questions about your company’s policies. A general AI model may be good at explaining what parental leave usually means, but it doesn’t automatically know what your company’s current policy says. If it guesses, you could end up giving someone the wrong answer with a lot of confidence.
One way to handle that is retrieval augmented generation, otherwise known as RAG.
For example
When someone asks a question, the system searches the documents you’ve provided, finds the relevant information, and gives that information to the model along with the question. The model can then use it to prepare an answer.
Why does that matter?
Simply because it lets you build applications around information that’s specific to a company, a customer, or a changing subject, all without training a new model every time a document changes. Instead it goes and finds the document isn’t enough on its own.
(Although you still have to check whether the system found the right information and whether the answer accurately reflects it).
AI Engineering: RAG for LLMs
If the document question-answering project in the LangChain course caught your interest, our AI Engineering: Retrieval Augmented Generation (RAG) for LLMs course takes you much further into this particular skill:
You’ll learn how to prepare information so the system can search it, retrieve material relevant to a question, and use it in an AI response. The course then puts that knowledge into practice with projects involving PDFs, customer feedback, and other sources of information.
This is the course to learn when your idea depends on the AI knowing something beyond its general training. RAG gives you a way to bring that information into the conversation, and the course shows you how to build and check that process.
Build AI agents that can use tools and complete tasks
We’ve already looked at building visual workflows with n8n. The courses below are for people who want to build and control agent behavior in code.
Build AI Agents with CrewAI
One way to organize a bigger task is to give different parts to different agents. For example, one could gather information while another examines it and prepares questions for you.
Build AI Agents with CrewAI introduces that approach using a framework called CrewAI. You’ll define the agents’ roles, give them tasks and tools, and connect their work into a process:
The course uses an AI interview coach as its hands-on example, so you can see what it means for several agents to contribute to one result.
Learn MCP (Model Context Protocol)
If an agent needs to do more than generate text, it may need access to a tool or source of information. Perhaps it needs to read documentation, look up a payment, or retrieve something from another service. Each connection needs a way for the AI application and the external tool to communicate.
Model Context Protocol, or MCP, provides a common way to make those connections. Our byte sized Learn MCP course introduces the idea and lets you work with MCP servers:
It’s worth learning when you’ve started building AI applications and want to understand how to give them controlled access to useful tools.
Build a ChatGPT Deep Research Clone with Streamlit
If you’d rather learn by building one complete example, our Build a ChatGPT Deep Research Clone with Streamlit course takes you through a research assistant that searches, works through the results, and produces a report with citations.
You’ll also give it an interface so someone else can use it:
This course is a good way to see how the ideas come together in one project.
AI Agents Bootcamp
Our AI Agents Bootcamp course goes wider, using Python and tools including CrewAI, LangGraph, MCP, and OpenAI’s Agents SDK to build several agent systems:
It includes some of the ground covered by the focused CrewAI and MCP courses, so you can choose the bootcamp if you want a broader route through agent development.
Learn to build and train AI models
Every course so far has involved using a model that already exists, but what if the problem you want to solve calls for working on the model itself?
That takes us into machine learning.
Instead of writing a separate rule for every situation, you give a model examples and train it to find patterns it can use on new data.
Complete A.I. Machine Learning and Data Science
If you want the broadest starting point in this area, our Complete A.I. Machine Learning and Data Science course takes you through Python, working with data, training models, and evaluating what they produce:
It’s a good choice if you want to understand the full process, from messy data, to a model you can test.
Machine Learning with Hugging Face Bootcamp
You don’t always have to start with an empty model. Often, you can take an existing one and adapt it to a more specific task. Hugging Face provides tools and a place to find datasets and models, work with them, and share what you’ve built:
In our Machine Learning with Hugging Face Bootcamp, you’ll prepare data, customize models, measure how well they work, and publish demos.
Choose this course if you want hands-on experience with the models and tools people can build on today.
PyTorch for Deep Learning Bootcamp
Deep learning uses neural networks, which are models made up of connected layers that learn useful patterns from data. PyTorch is a software framework that helps you build, train, and test those networks.
PyTorch for Deep Learning Bootcamp goes into that process in depth. You’ll work from the data and the numbers a model can process through to training a neural network and using it on real problems:
This is for you if you want to understand and build deep learning models, rather than only call an existing model through an API.
TensorFlow for Deep Learning Bootcamp
TensorFlow is another framework for building deep learning models. Our bootcamp covers applications such as recognizing images, working with text, and forecasting from data over time:
Work with data and adapt models for specific problems
Once you start building models, you run into a practical problem, which is that the data rarely arrives in a form you can use straight away...
(Typical right?)
And even when a model already exists, it may need to be adapted or connected to other parts of a system before it can solve your particular problem.
That's why these next courses focus on different parts of that work. You wouldn’t necessarily take all three. The right one depends on whether you want to prepare data, customize a language model, or work with images.
Data Wrangling Bootcamp
Imagine you’re trying to predict which customers might cancel a subscription. You have records from several systems, but some dates use different formats, some entries are missing, and the same customer appears more than once. If you train a model on that data as it stands, those problems can affect what it learns.
Data wrangling means exploring, cleaning, and reshaping data so you can analyze it or use it to train a model. Our Data Wrangling Bootcamp teaches you to do that with Python and Pandas, starting from the basics.
You’ll work with messy datasets and build toward a customer churn prediction project:
It’s worth considering this course if the machine learning courses sound interesting but you’re wondering what happens before the model gets its training data.
AI Engineering: Fine-Tuning LLMs
We’ve already looked at RAG, which finds relevant information and gives it to a model when someone asks a question. Fine-tuning does something different, because you continue training an existing model on carefully prepared examples so it becomes better suited to a particular kind of task.
For example
If you need a model to consistently classify a specialist type of document, you might test whether fine-tuning helps it learn the patterns in labelled examples. You’d still evaluate the result carefully; more training doesn’t automatically make a model better.
AI Engineering: Fine-Tuning LLMs takes you through preparing a dataset, adapting an open model, and deploying it:
This is a more specialized choice for someone who wants to work directly on model behavior, rather than simply supply documents to an existing model.
The Computer Vision Bootcamp
What if the information your system needs to understand is an image? Perhaps you want to find an object in a photograph or identify which part of an image matters.
Well, our Computer Vision Bootcamp explores how vision models process images, including models that can separate an object from the background around it. You’ll work with those models and build a vision pipeline that can be deployed on AWS:
This is a good course to take if images or video are central to the problem you want to solve.
Build and run AI systems on AWS
A project can work perfectly on your own computer and still have a long way to go before other people can use it. You need somewhere to run it, a way to handle requests, and a plan for what happens if the system is slow, expensive, or starts producing poor results.
That’s why some of our courses move from building an AI feature to running it on cloud infrastructure.
AWS Bootcamp: Build AI Apps with AWS Bedrock
If you want to build an application using existing AI models, AWS Bedrock is Amazon’s platform for working with them through AWS. You can use it to build features such as document summaries, image generation, and AI-assisted workflows without training the underlying model yourself.
Our Build AI Apps with AWS Bedrock course takes you through building and deploying several projects. It also covers evaluating and monitoring model output:
Choose this if you understand the idea of calling an AI model from an app and want to learn how to build those features within AWS.
AI Engineering: Build, Train, Fine-Tune and Deploy Models with AWS SageMaker
SageMaker is useful to learn when you want to work more directly on training, customizing, and deploying models. This course combines that with both Hugging Face and Pytorch:
You’ll work with models, test their results, deploy them, and explore how to monitor and scale them. It’s the more relevant choice if the model development sections interested you and you’re ready to learn what it takes to make that work available beyond your own experiments.
Build AI Agents with AWS
What if the system you want to deploy is an agent rather than a single model-powered feature? It may need to use several AWS services, call tools, and coordinate parts of a task.
Our Build AI Agents with AWS course teaches that through an AI travel agent project:
You’ll build and deploy an agent workflow using AWS services and see how several agents can work together.
AI Voice Agents with AWS
Voice adds another challenge because the system needs to handle spoken input and respond quickly enough for a conversation to feel natural. Delays that are barely noticeable in a text chat become frustrating when someone is waiting for an answer out loud.
Our AI Voice Agents with AWS course focuses on building that kind of real-time interaction with AWS Bedrock:
It covers the audio flow and tool connections behind a voice agent. You’d choose this for a voice-based product or project, rather than as a required step for every AI developer.
Explore other ways to build and run AI systems
AWS is one place to build an AI system, but it isn’t the only environment you might work in. Some teams use Microsoft’s Azure platform. Others want to download a model and run it on infrastructure they control.
These next courses help you understand those options.
Introduction to Microsoft Foundry
Microsoft Foundry is an Azure environment for building AI applications and agents with models and services from different providers.
Our Introduction to Microsoft Foundry course starts with the building blocks, so you’ll see how the environment is set up, how an agent can communicate with Azure services, and where tools and human decisions fit into a workflow:
It’s a useful introduction if your workplace uses Azure or you want to explore AI development in Microsoft’s ecosystem before committing to a larger project.
Azure Bootcamp: Build AI Workflows & Agents with Microsoft Foundry
Suppose a customer support request comes in. An AI system might look up relevant information, prepare a draft reply, and send it to a person for approval. To build that well, you need to decide which steps follow fixed rules, where an AI model makes a judgment, and when a human takes over.
Azure Bootcamp: Build AI Workflows & Agents with Microsoft Foundry gets you building those kinds of processes:
You’ll create a customer support agent and a workflow that keeps a person involved in an important decision.
Open Source Models
So far our examples have involved sending a request to a model run by a provider. However, you can also explore models that you can download and run yourself. But that raises new questions like:
Which model is suitable for your task?
Will it run on the hardware available to you?
How do you tell whether its answers are good enough?
Etc
Our Open Source Models course helps you figure this all out:
You’ll compare models using tests relevant to your own needs, then set one up locally with LM Studio. It’s a good choice if you’re curious about model choice and want hands-on experience beyond using the familiar chat tools.
Local AI Revolution: Ollama and OpenClaw
Running a model on your computer is a start, but what if you want to run it on a server you control and connect it to an interface or an agent?
Our Local AI Revolution: Ollama and OpenClaw course walks you through that setup:
You’ll use Ollama to run a model, add a chat interface, explore using a rented GPU for more computing power, and connect an agent to the system.
This is also handy if you want to learn about hosting and controlling the infrastructure yourself. However, running your own setup also means you’re responsible for maintaining it, so it’s a different commitment from opening a hosted chat app.
Keep an AI system working after launch
Putting a model online isn’t the end of the job. New data arrives, people use the system in ways you didn’t expect, and its results may change over time.
For example
Imagine a model that helps sort customer support requests. It works well when you first test it, but six months later your company has new products and customers are asking different questions.
How would you notice if the model was sending more requests to the wrong team?
And if you trained an improved version, how would you check that it really is better before replacing the one people use?
That work is called MLOps. It brings together machine learning and the systems used to deploy, monitor, test, and update models. The aim is to make improvement a process you can manage, rather than discovering a problem only after it has affected users.
MLOps Bootcamp: Build Real-World AI Infrastructure
MLOps Bootcamp takes you through that process with an image model running on AWS. You’ll train and deploy it, monitor its predictions, collect new examples, and create a way to retrain it. Crucially, the course also covers checking a new version before it replaces the one already in use.
This course makes sense if you’re interested in what happens after a model has been built, particularly the engineering needed to keep it useful as its data and users change:
If you’re still learning how to train and evaluate your first model, start there and come back to MLOps when you’re ready to run one over time.
Understand why AI behaves the way it does
You can get useful results from AI without knowing every detail of how it works. But as you use it for more important tasks, you’ll start asking harder questions:
Why did the same prompt give you two different answers?
Why did the model follow one instruction and ignore another?
And how can you tell whether a prompt that worked once will work reliably?
These courses help you investigate those questions. They’re a good fit if you enjoy experimenting with AI, or if you need to choose, test, and use models with more confidence.
See what’s happening around the model
When you use ChatGPT, you’re interacting with more than a language model. There are instructions, memory, and tools that affect what it can do and how it responds.
Our ChatGPT Deep Dive walks you through those pieces, including how models can work with images and audio and call tools to complete tasks:
If you’ve wondered why ChatGPT can remember a preference or fetch information that isn’t in your prompt, this course helps you make sense of it.
Then there are the settings that affect a model’s output.
For example
You might want a wider variety of ideas for a brainstorming task, but more predictable responses when you’re extracting information into a fixed format. Our LLM Hyperparameters course introduces controls such as temperature and Top P, and shows you how changing them can affect the answers you get:
Go beyond basic prompts
Once you can write a clear prompt, the next challenge is handling tasks that take several steps.
For example
Suppose you want AI to read customer feedback, group similar complaints, and suggest which issue to investigate first. You could try asking for everything at once, but you might get a better result by breaking the work into stages and checking what comes out of each one.
Advanced Prompting Techniques explores approaches such as prompt chaining, ReAct, and Tree of Thoughts. You’ll experiment with how to structure more demanding tasks and learn where those techniques came from:
It’s a useful next step if you already know the prompting basics and want more ways to tackle problems that resist a simple question.
You might also notice that some models seem to spend more effort on a difficult question before answering it. AI for Beginners: Reasoning Models explores how these models behave, how they’re trained, and where their apparent reasoning can still mislead you:
That matters when you’re deciding whether to use one for a complex task and how carefully to check its answer.
Test what works and spot what can go wrong
Let’s say you’ve written a prompt that summarizes support tickets. It works beautifully on the three examples you tried. But will it still work when a ticket is vague, unusually long, or written in another language?
Applied Prompt Engineering: Prompt Testing & Model Benchmarks teaches you to build test cases, compare prompts and models, and evaluate the results. Instead of deciding that a prompt “feels good,” you can check how consistently it handles the situations you care about.
Reliability also means understanding failures...
The Dark Side of AI looks at hallucinations, prompt injection, data leakage, and other ways an AI system can go wrong. If you’re putting AI into a workflow or product, you need to recognize those risks so you can decide where to add checks and human review:
Finally, if you’re curious about the bigger questions behind these tools, AI Research & The Quest for Artificial General Intelligence introduces ideas such as model scaling and research into what happens inside neural networks.
It gives you a way to follow new AI developments with more understanding, even if you don’t plan to become a researcher.
Apply AI to a problem you actually care about
By now, you’ve seen the main skills behind working with AI such as prompting, coding, connecting models to information, building agents, and training models. But you don’t have to learn them in isolation.
If there’s a particular kind of work you want to do, a focused course can give you a problem to solve while you practice.
Use AI for research and client work
Say you’ve been given a stack of reports and need to find the points that matter for a meeting. The NotebookLM Guide shows you how to work with an AI assistant around material you provide, so you can explore the documents and organize what you find.
If you regularly need to investigate a topic more broadly, Perplexity AI for Professionals focuses on using AI for research and decision-making. In either case, the useful skill is knowing what question you’re trying to answer and checking the source material before you rely on the result.
If you freelance, the problem may be less about research and more about winning and delivering work. Enhanced Freelancing with AI looks at using AI to improve your profile, develop proposals, and manage client projects:
You could, for example, give it the details of a client brief and ask it to help you identify questions to raise before you write your proposal. You’d still decide what to promise and how to do the work, but you’d have help getting from a blank page to a useful first draft.
Explore AI in cybersecurity
AI creates opportunities for security teams and new things they need to watch for. An attacker might try to hide malicious instructions in a document that an AI assistant reads, while a defender might use AI to help examine suspicious activity.
If you want an introduction to that relationship, Artificial Intelligence in Cybersecurity covers the risks and vulnerabilities associated with AI systems:
If you already want to practice security work, AI for Ethical Hacking & Cybersecurity Bootcamp goes further with hands-on scenarios and tools. This one helps you understand what AI changes in cybersecurity; the other helps you work through those changes in practice:
Build a project in a field that interests you
Perhaps you already know you want to build with AI, but an abstract demo doesn’t hold your attention. Pick a project whose purpose you can picture.
If customer support interests you, Build a ChatBot with Nuxt, TypeScript, and the OpenAI Assistants API walks you through a chatbot for a particular product or service:
You’ll have to think about what customers ask, what information the bot needs, and what a useful answer looks like.
If you’re drawn to finance, Build an AI Stock Analyzer using ChatGPT, Python and LangChain gives you a shorter project focused on gathering and examining stock data:
Vibe Code a Generative AI Finance App with Python and LangChain takes that interest into a larger app-building project:
In both cases, the point is to practice turning data and AI tools into something a person can use, while learning to question the output rather than treating a generated answer as a fact.
And if you want a first project without taking on a coding course, Build an AI Career Coach using an Open Source LLM shows you how to make a coach that can teach, quiz, and challenge you using natural language:
It’s a way to put your prompting skills to work on something personal before deciding whether you want to learn to build more technical AI applications.
You don’t need to take every course here. The one that’s easiest to stick with is often the one that helps you solve a problem you’re already curious about.
So what should you learn next?
So as you can see, there's so many different ways to upskill for AI. You could get better at using it in your current work, use it to help you write software, build an app or agent, or learn to train and run the models behind it.
It's easy to be overwhelmed, when you see all these options, but don't freak out because you don't need to learn them all!
Just start with one thing you’d like to be able to do.
And if you’re still deciding, start with our AI for Beginners path.
Become a AI For Beginners
16 milestones 16 courses
Step-by-step roadmap where you'll learn to code and build a portfolio.
Curated curriculum of courses, workshops, challenges, projects, and action items.
Become a AI For Beginners from scratch and actually get hired.
Earn on average per year:
---
US salary data collected from Indeed, LinkedIn, and Web3.career 2026.
It’ll give you a foundation you can use whichever direction you take next.
Best articles. Best resources. Only for ZTM subscribers.
If you enjoyed this post and want to get more like it in the future, subscribe below. By joining the ZTM community of over 100,000 developers you’ll receive Web Developer Monthly (the fastest growing monthly newsletter for developers) and other exclusive ZTM posts, opportunities and offers.
No spam ever, unsubscribe anytime
