How to Get Your First Job in Data Science or AI in 2026

I get a lot of LinkedIn messages from people trying to break into data and AI. I try to respond to all of them. At the peak, I had 2-3 people I was actively mentoring with bi-weekly check-ins. Over the years, I've watched dozens of people go through this process. Some land roles in a month or two, others struggle for much longer.

More than half stop responding after I give advice. It could be my advice comes across as harsh. Rather than commiserating about external factors, I try to focus on what's controllable. Here's what I've learned from watching this play out.

A caveat before we start: if you're an international fresh graduate trying to break in, it's significantly harder and most of it is out of your control, especially with the current political climate. I don't have great advice there other than to acknowledge that it's seriously hard, and that really sucks.

The Bar is Rising Fast

I must say, I was confident in my advice until recently. Times have changed, and advice from someone who recently went through the grind might be more relevant than mine. The patterns I observed over the last 6 or 7 years have become unpredictable given the economic and political situation, plus advancements in AI. Because of this, I actually respond a bit less to LinkedIn messages.

When I got my first data science job around 2016, the bar was pretty reasonable: an engineering master's degree and the ability to code. I started doing research at the US Geological Survey, writing scripts that automated some simulations.

The bar has changed a lot since then. This is super rough and debatable but I think gets the point across:

Era	What Made You Stand Out
2016-2018	Can you code and learn ML algorithms (engineering background)?
2018-2021	Deployment: Docker, cloud, CI/CD, observability
2021-2023	Modern data stack: dbt, Snowflake, orchestration
2023-2024	A RAG project
2025	Agent implementation

Notice how the cycles are getting shorter? It used to be 3-year windows. Now it feels like 1 year or less. I'm not entirely sure what I'd recommend as a portfolio project right now. A year ago, building your own RAG chatbot was impressive. Today, that's trivial to spin up. Anyone can spin up an elegant chatbot website in half a day.

What does impress? Projects that show you understand the hard parts: evaluation, reliability, cost optimization, edge cases. Anyone can build a demo. Fewer people can explain why their system fails and how they'd fix it. If you're building an agent, show the failure modes you discovered and how you handled them. If you're doing RAG, show your retrieval evaluation metrics and how you improved them. The bar isn't "can you use the API", it's "do you understand what happens when things go wrong."

A degree or bootcamp graduation is really just the starting point. The learning doesn't stop. I've seen senior people struggle to land roles post-layoff because the landscape shifted while they were heads-down at their last job. The people who thrive tend to be genuinely curious and keep learning on their own. When someone isn't passionate about this stuff, it shows.

That said, I don't advocate non-stop grinding. The field is competitive because it's currently sexy. It doesn't come easy anymore. But I do believe it just takes one or two sprints to get a leg up.

The 1 Standard Deviation Rule

Here's a way to think about it: if your resume fell in a pile of 10 applications, what are the chances yours would come on top? If you're average, chances are pretty bad because there will be 4-5 people ahead. But if you invest just a little bit of time to improve, you can be one standard deviation above the mean. That's the 84th percentile. That increases the odds tremendously, and it's totally achievable. After that, it might be a bit of diminishing returns.

Most people settle for median. They do the bare minimum - take a course, update their resume, and start applying. A bit more effort goes a long way.

We're in the golden age of tutorials. There are so many options that you can find an instructor whose style resonates with you. That's amazing. But here's the thing: all your peers have access to the same resources. Everyone's taking the same courses, following the same tutorials. The people who stand out are doing something beyond the standard curriculum.

I often find job seekers are unaware of where their skills sit relative to the average. It's like the study where up to 80% or more rate themselves as above average drivers. When applying for the first role, most people haven't seen what the rest of the market looks like. A lot of job seekers think they're already okay when there's still room for improvement.

Ability vs Visibility

There's a difference between ability and visibility. You need both to get a job, but a lot of people get this backwards. I see folks grinding leetcode before they've even gotten a single interview. If you're not getting interviews, the problem probably isn't your coding skills. It's your visibility.

Think about it: between a candidate who's an amazing coder but has no public portfolio, versus a decent engineer with a great portfolio, the one with the portfolio is more likely to get the job. They're visible. Recruiters can find them. Hiring managers can see their work.

In my experience, ability only becomes the bottleneck once you're getting interviews but failing them.

Gaining Visibility

The people I've seen break in tend to pick a lane and own it. A few patterns I've noticed:

1. Chase the frontier - Some people stay on top of the latest AI trends: agents, new frameworks, cutting-edge techniques. They build projects with bleeding-edge tools and write about what they learned. They post experiments on GitHub with good READMEs, write threads or blog posts breaking down new papers.

2. Go deep in a tech niche - Others become the Spark expert, the PyTorch specialist, the Kubernetes person. Less glamorous but consistent demand. They write tutorials that solve specific, painful problems.

3. Domain specialization - Others focus on biology, real estate, sports, finance—combining tech skills with deep domain knowledge. This is powerful because they're not competing with pure ML engineers. They become the rare person who understands both the data and the business.

The common thread I've noticed is the ones who succeed create artifacts. A GitHub repo with a good README, a blog post that you can link on your resume. Speaking at conferences or meetups is also more accessible than most people think. The bar to get a talk accepted at a local meetup is pretty low. But slop won't help. It's immediately visible to practitioners when a blog post is AI-generated filler with no real insight. It needs to be something you're actually proud of, something that shows you've thought deeply about the problem.

One caveat: some domains like quantitative finance or biotech often require advanced degrees or prior industry experience, making them harder to break into from the outside. I've seen people get around this by targeting tangential roles first to build relevant experience.

Conclusion

There's no magic bullet here. Supply has outpaced demand. The bar is higher than it's ever been.

One mindset thing I've noticed: it's tempting to think "man, these people who got in earlier aren't even that good." And honestly, sometimes that's true. But it's not a useful frame. The competition isn't people who already have jobs—it's other applicants for the same role.

The bar is higher, but still well within reach.