The Exact ML Project I’d Build to Get Hired in 2026

over 100 applications and portfolios, and only a few times has someone’s project wowed me enough to progress them to the interview stage.

So in this article, I’m going to give you the exact framework I developed and followed to find your perfect ML project that will land you a job.

Let’s get into it!

Why Most ML Projects Fail

Let me tell you something that all hiring managers think across every company I have worked for.

When we see a house price prediction model or a Titanic survival classifier, we don’t think “solid fundamentals.” We think “next.”

I am not even joking.

These projects have been done so many times that they don’t tell me anything about the person.

It only tells me that they can follow along a bog-standard tutorial and replicate the results.

A project that gets you hired has four key things:

• It’s personal — You genuinely care about what it’s predicting.
• It’s novel — I haven’t seen it a hundred times before.
• It’s relevant — It connects to the kind of work you want to be doing.
• It’s live — People can actually see it in action.

Get all four right, and your project makes the hiring manager remember you, and this is coming from my personal experience of hiring.

The problem is nobody can hand you a project like that. It has to come from you.

So instead of handing you an idea, I’m going to give you a framework to follow to develop a project like this.

I have also turned this framework into a 7-page downloadable PDF workbook you can check out in the description below to work through and find the right project that will get you hired.

Example Project

Before I dive into the framework, let me give you an example of a project from a candidate that we hired.

At one of my previous companies, we were hiring for a junior data scientist to work on optimisation and operations research problems.

The candidate we hired stood out for one main reason: they had a highly relevant and deeply personal project that closely matched the role.

They were passionate about NFL fantasy football and wanted to improve the way they built their weekly team selections.

So, they developed their own optimisation engine to allocate players more effectively within the program’s constraints.

It wasn’t just the engine itself; they read academic papers on optimisation strategies and studied how others were approaching the same problem.

This project hits on all four bullet points we mentioned earlier:

• It’s personal – It was a personal problem that they were interested in.
• It’s novel – It was unique, and we hadn’t seen anything like it before or since.
• It’s relevant – it showed their passion and interest in optimisation and operations research, which was exactly what we were hiring for.
• It’s live – It was directly relevant to the job for which they were applying.

Let me now break down the exact framework you can follow to build a project exactly like this.

Start With Your Interests

When people look for a project to build, they open a list of ML datasets, most likely in Kaggle, and try to find something interesting.

That’s backwards.

Start with yourself and your interests.

More specifically, write down five things you genuinely care about outside of work and outside of data and ML.

Focus on your hobbies, obsessions and other things you would happily talk about for an hour with no problem.

When I did this, my list was something like:

• Investing
• Hockey
• Gym/fitness
• Films
• Youtube

Why do we need to pick something that interests us?

Because a project you care about and are interested in is a project you will finish.

I can’t stress enough how much easier it is to do a project that genuinely motivates you rather than one you “think” you should do.

Once you have your five interests, please write five questions for each interest that you genuinely want answered.

For example, “Which fantasy football players are underpriced this week?” is a question, whereas “Football stats” is not.

Don’t overthink it and simply write things down.

You will now potentially have 25 project ideas, that are very likely to be completely unique or at least not many people have seen.

Filter Your Top Picks

Now we need to cut that list down to our top picks.

The first step is to remove the questions or ideas that are not obviously ML problems. For example, “Why do I enjoy films?” is a great question, but it’s not a machine learning project.

Machine learning at a really high and crude level can be broken down into 5 key areas:

• Predicting a number — regression.
• Predicting a category — classification.
• Forecasting over time — time series.
• Recommending things — recommendation systems.
• Grouping things together — clustering.

Go through your 25 or fewer questions and find the ones that fit one of those 5 areas and remove the ones that don’t.

This should leave us with around 10-15 feasible ideas that we can solve using ML.

Now you need to pick one, and the way to do this is to individually score these ideas against the following criteria:

• How personal is it?
• How novel is it?
• How relevant is it for the roles I am going for?
• How hard is it to get the data?
• How hard is it to build?

Rate each one out of 5, sum them all up, and the one with the highest score is the one you will build.

Validate The Project

Before you commit weeks to building this project, I want you to run three quick checks.

First — Where exactly does your data come from? Name a real source — an API, public dataset or any other unique source. If you can’t name one, finding the data is your first job.

Second — Could you get a rough first version working in about two months, assuming an hour or two a day? If it’s bigger than that, shrink it. A small project you actually finish beats a large one you abandon — every single time.

Third — How common actually is it? Nothing is ever truly original, but if this project is something you have seen several times before, then maybe reconsider it and pick your second choice.

Pass all three checks, and you’re done.

Well done! You’ve got a project that’s yours, that a hiring manager hasn’t seen a hundred times, and that you can actually finish.

Make It Live

Like most people, you will probably do the initial research and prototyping of this project on a Jupyter Notebook.

However, companies nowadays want people who can deploy their solutions to create business impact.

Even if your model is the best thing since the transformer, it’s useless if it is stuck inside a notebook.

Deploying a model is actually not as complex as people think. I’ve coached several of my clients to build their first end-to-end machine learning model with no previous experience using the following tech stack and process:

• Build the prototype solution in a Jupyter Notebook.
• Break down that Jupyter Notebook into individual Python files that follow production code standards, implementing features like typing, formatters, and docstrings.
• Add your Python files to a git repo and create a great README that explains the project.
• Add all the key software engineering tools and concepts, including unit tests, Poetry dependency management, Makefiles, and PyEnv.
• Create a Streamlit dashboard to showcase your results, and deploy it on the Streamlit community cloud.
• Set up your repo to run daily using GitHub Actions.

Bish. Bash. Bosh.

You have just deployed your model end-to-end using industry-standard tools, which I have been using as a machine learning engineer at top tech companies for years.

I can appreciate that it may seem completely overwhelming for someone brand new to this, with no one by your side to guide you, so I have created a template repo with all the boilerplate code to set this up.

Another Thing!

Connect With Me