If you’re studying a field like data science, software engineering or machine learning, the start can feel overwhelming. Often, it’s not clear where to start, and it’s not clear how to build a portfolio of projects.

If you’re lost in what you should be doing, try the following technique.

Pick any single topic in your field that interests you. Learn everything about that topic. Publish the code and notes that you made during the learning process.

Why is this a good process? Because when you do something in depth, you go further than most other people are willing to go, gaining unique skills. You will both learn a skill, and show potential employers that you went through the process of learning that skill.

What if you’re stuck choosing?

Just choose - ask a friend, roll a dice, choose a bulletpoint from the skill list of your dream employer. Don’t get stuck. When making a portfolio, one day spent coding is much better than one day thinking about what to code.

It’s not about making the perfect choice. If you choose a bad topic you will still gain the understanding of why it’s bad. The main goal is to put in the work, moving one step forward. You will gain technical proficiency, and be better at making your next choice.

On why a single step forward is a good idea, check out Alex Komoroske’s essay on iterative adjacent possible.

What if it’s hard to learn everything about the topic?

Let’s assume that you’ve chosen AlphaFold as your topic. If your goal is to learn everything about AlphaFold, you might discover that there is too much to learn. Decompose the problem and choose a simpler topic that AlphaFold is build upon.

Here are two ways you could decompose a larger topic:

• ask ChatGPT: “What should I learn to understand AlphaFold?”,
• read any text about AlphaFold, and as soon as you don’t understand one thing, choose that as your next topic.

For example, reading the Wikipedia entry, I can see that I don’t understand what does it means to predict protein structure. Then, my next topic of study is protein structure prediction. Again going to wikipedia.

If you cannot write code that captures your understanding, then it’s likely that you don’t understand a topic completely - dig in!

How do I publish code and notes?

Here is a very simple method. While learning, write down all links that you visit and why you visited them in a README file. If you write code - or even just run code from somebody else, write down the steps it took. Publish that README on Github.