Given the wide range of backgrounds interested in AI these days, I’ve tried to make this guide widely accessible. However, no guide can help everyone. Here are a few specific groups I have in mind.
- Technical professionals trying to up-skill for career advancement
- Business leaders who haven’t coded in years, trying to keep up with the changing tech landscape
- Entrepreneurs building AI-native products
- Students trying to develop their technical AI skills
About me — I’ve worked in AI for the past 6 years. I started as an AI researcher while getting my PhD, then eventually worked as a data scientist at Toyota. Although I still have a lot to learn, the approach below covers (what I think are) the essentials based on my personal experience.
The guiding principle of this framework is to learn by doing. Each step outlines a clear and specific objective through which completion will naturally develop key skills. In other words, rather than reviewing a list of concepts and courses, each step is a task designed to force me to learn essential skills by completing it.
Here’s an overview of the 5-step approach. Each step builds upon the ones before it.
- Use ChatGPT (or the like)
- Install Python
- Build an Automation
- Build an ML Project
- Build a Real-world Project
If starting from zero, the first thing I would do is familiarize myself with modern AI tools i.e. ChatGPT, Claude, and the like. This is important because frequently using these models will give me a practical understanding of what they can and can’t do and develop my ability to use them effectively through prompting.
On a more meta level, these chat interfaces are incredible tools for learning AI (or anything else, really). I’d use it to explain confusing buzzwords and technical concepts (e.g. LLM, tokens, API, RAG) and be sure to ask follow-up questions until I have a solid understanding of each idea. For those that don’t click, I’d seek alternative resources using Google search and YouTube.
Although I could go far with today’s no-code AI tools, they are fundamentally limited. Namely, these tools can’t be easily used to build custom solutions or process information in bulk. That’s why the next thing I would do is install Python on my computer.
Python is the industry standard programming language for AI development. To get it installed, I’d ask ChatGPT for step-by-step instructions. If I get stuck, I’d come back to ChatGPT, explain the issue, and ask for additional guidance.
While using ChatGPT (or any other AI assistant) in this way can streamline the process significantly, I would still take the time to understand each step of the process and ask follow-up questions as needed. This is an important habit to develop because it will avoid accumulating technical debt, which I’ll have to pay later when something goes wrong.
Once I’ve become comfortable using ChatGPT and installed Python on my machine, my next step would be to build a simple automation using Python. My approach to generating project ideas would be to think of things I consistently use ChatGPT for (e.g. summarizing research articles), then try and automate it with Python.
This would require me to become familiar with OpenAI’s Python API. So, I’d start by reading their documentation and reviewing the example code there. Once I felt comfortable with the API, I’d start writing Python code.
My first step would be to think through the steps of my automation. For example, if summarizing research papers, the steps might be:
- Read paper contents into Python
- Construct prompt for GPT-4o
- Make an OpenAI API call
If I got stuck, I’d turn to ChatGPT for assistance. For instance, if I didn’t know how to read PDFs into Python, I could ask ChatGPT for help. If it spits out code I don’t understand, I’d ask follow-up questions until I understand each line.
It (again) is important that I take this approach to coding with ChatGPT because blindly copy-pasting code from it wouldn’t teach me much. It would also accrue unforgiving technical debt. In other words, I’d get short-term gains but would have to pay for them later via technical difficulties and headaches.
Task: Use OpenAI API (or the like) to build a simple automation
Resources: OpenAI API Intro | Paper Summarizer Example
After Step 3 gets easy for me, I’d seek out more sophisticated projects. Rather than simply making ChatGPT-like API calls, I’d build a project that required me to use embedding models or to train a model myself.
Potential project ideas would be things like:
For example, if I went with the RAG project, I’d first educate myself on RAG by watching YouTube videos and reading blog posts. Then, I’d break down the system’s basic components and the steps to implement it. Finally, I’d start coding the project, using ChatGPT as a co-pilot like Step 3.
- Task: Build an ML project that goes beyond ChatGPT-like API call
- Resources: More Project Ideas
Although I would have learned a lot about the technical side of AI from Steps 3 and 4, this is not sufficient for generating value with it. For that, I’d need to use what I learned to solve real-world problems.
There are two ways to do this. I could, one, solve my own problem, or two, solve someone else’s problem. Since I (hopefully) already did the former way in Steps 3 and 4, here are a few different ways I’d approach the latter.
- Reach out to business owners and professionals in my network
- Join a research group at my University (if I was a student)
- Find an internship (if I was a student)
- Find a freelance gig on Upwork
Let’s say I had graduated from college and wasn’t quite confident enough to freelance yet, so that leaves Option 1. I’d start by making a list of people to reach out to. Ideal contacts would be small business owners or professionals working at a small to medium-sized business.
Then, I would craft a message like the one below and send it to everyone on my list via LinkedIn DM or email. If I struggle to find the right wording, I’d use ChatGPT (yet again) to help out.
Subject: Offering Free Help with AI ProjectsHi [Name],
Your work at [Company Name] caught my attention—[insert a specific detail or
observation, e.g., “it’s clear you’re doing innovative things in X” or “your
focus on Y stood out to me”].
Over the past few months, I’ve been building practical AI projects to develop
my skills. You can see some examples [here](link to portfolio).
Now, I’m looking to apply my learnings to solve real-world problems
by helping businesses like yours—**completely free of charge**. If there’s a
challenge you’ve been looking to automate or improve with AI, I’d be happy to
explore how I can contribute.
Would you be opposed to a short conversation to discuss this?
Best regards,
Shaw
Although AI entails an interdisciplinary collection of technical skills and knowledge, with today’s tools and resources, it’s never been more accessible. Here, I shared the 5-step approach I’d take to learning it today.
That said, it’s important to remember that learning (itself) is hard. You will get confused, you will get frustrated, and you will question why you’re putting yourself through this. However, if you are willing to see it through, you will be rewarded with clarity and knowledge, which is an amazing gift.
If you have questions or want feedback on project ideas, feel free to share them in the comments 🙂