The world’s leading publication for data science, AI, and ML professionals.

How To Generate Machine Learning Use Case Ideas For Your Portfolio Project

Three Idea Creation Techniques

Photo by Mark Fletcher-Brown on Unsplash
Photo by Mark Fletcher-Brown on Unsplash

Your portfolio project could make or break your chances of landing a job. With a growing number of people aspiring for a career in machine learning, it’s important that you’re able to distinguish yourself from the other candidates. In this light, many use their portfolio project to set themselves apart and grab the attention of hiring managers.

You can work on competitions from Kaggle for your project. But, realizing the full experience of an ML engineer comes from working through the entire machine learning workflow, which is best done when working on your own unique project – Kaggle skips the early stages of the machine learning workflow. Although not impossible, generating potential use cases can be tough.

Below are three techniques that would help you generate potential use cases for your next project.

Use The Magical Island

Machine Learning is a highly technical domain. Thinking about all of the technicalities when you’re trying to come up with a fun use case is a great recipe for disaster. We end up thinking about a highly complex use case that may take a team of six people, twelve months to develop – precious time that we don’t have when we’re looking for work.

Instead, the simple solution is to use Cassie Kozyrkov‘s magical island. For those unfamiliar with it, Cassie proposed an island in the middle of the ocean where many of her friends work. Their job is to come together and pretend to be AI – and you get to keep all the credit for their brilliance (or stupidity).

For this magical island to work, you must send them an input. The workers will do whatever they do, then send you back a decision. However, you’re not allowed to give them any rules, they must learn with examples. Your job is to give them work to do. The easiest way to find work for them is by casting off repetitive tasks.

What repetitive drudgery would you offload?

The answer to this question is one step closer to finding a plausible use case. List as many as you can:

  • Input: Property information, Output: Recommended Price
  • Input: Movie information, Output: Will I like it?
  • Input: Service request, Output: The amount of waiting time
  • Input: Photo at the zoo, Output: What animal is in the picture?
  • Input: Stock information, Output: The price of the stock tomorrow

Before you can fully invest yourself into one of the ideas, think about what it means for your workers to perform the task successfully – after all, they may be drunk off their face. You can’t trust them blindly!

"A common mistake businesses make is to assume machine learning is magic, so it’s okay to skip thinking about what it means to do the task well." – Cassie Kozyrkov, Advice for Finding AI Use Cases.


Read Other Use Cases

Learning from others will help you build your intuition. When I played football, I’d study Messi. I’d analyze how he dribbled past opponents, how he creates space for himself with and without the ball, and visualize myself in his shoes whilst he plays.

Of course, I didn’t become Messi, but it helped.

Whenever I found myself in similar scenarios on the pitch, I’d instinctively perform maneuvers without thinking, and it paid off very often.

In the same way, reading tons of machine learning use cases develops your intuitive eye. Eventually, your ability to spot machine learning use cases will significantly increase – as well as your ability to deal with various problems that occur during deployment.

"To learn to design machine learning systems, it’s helpful to read case studies to see how actual teams deal with different deployment requirements and constraints." – Chip Huyen, Machine Learning System Design

A number of companies share their machine learning use cases. Here are a few you could use to develop your intuition:


Strengthen Your Idea Muscle

Maybe your idea muscle is weak. While both of the aforementioned techniques will help you to generate ideas, it’s beneficial to deliberately strengthen your idea muscle.

Just like your biceps and quads, your idea generator needs to be trained to churn out ideas, regularly.

The "workout your idea muscle 10x a day" theme has been floating around Medium for some time – Ayodeji Awosika and James Altucher being popular advocates of the method.

Writers use the technique to consistently churn out unique ideas for potential blog posts, but it can be used to generate ideas for absolutely anything.

All it takes is a pen, paper, and a few minutes at the beginning of the day – or you can break it up. But you must write down 10 before the day ends.

In your case, you won’t be writing blog ideas. Your task would be to write ten machine learning use cases each day. Here’s the cool thing…

NOBODY HAS TO SEE IT.

If the ideas you come out with turn out to be trash then so whaT? Nobody is going to see it. All you’re trying to do is exercise your idea muscle with some resistance training.

Over time, your idea muscle will develop, making you into a machine learning use case generating machine.

Final Thoughts

Having a unique project is a great way to distinguish yourself from the pack. Many often struggle to generate ideas when thinking of things they can implement. Some techniques to help with idea generation include visiting Cassie’s island, reading other people’s use cases, and strengthening your idea muscle.

Thanks for Reading!

If you enjoyed this article, connect with me by subscribing ** to my FRE**E weekly newsletter. Never miss a post I make about Artificial Intelligence, Machine Learning, and Freelancing.

Related Articles

5 Factors That Contribute To Successful Machine Learning Projects

3 Signs You Don’t Need Machine Learning

How To Design A Machine Learning System


Related Articles