How to get into data in 2027

I get this question more than any other, and the honest answer changed in the last two years. The door into data is narrower than it was in 2021, the AI tooling changed what “junior” means, and most of the advice online is from people who got in when it was easier. Here is what I would actually do, by starting point, with the numbers on where entry-level jobs really are.

Entry-level responses by title

Data Engineer
55 · $90k med
Data Analyst
50 · $60k med
Data Scientist
38 · $80k med
Research Scientist
15 · $131k med
ML Engineer
13 · $83k med
Data Manager
2 · too few
Analytics Engineer
1 · too few

174 entry-level responses out of 3,104 in the full sample. Same source as the salary map: ai-jobs.net, 2022 to 2023, full-time. Medians hidden under five responses.

Start with where the jobs are

Filter the salary survey to entry level and count by title. Only about one in sixteen responses is junior at all.

Two doors

US only, same picture: analyst and engineer hold most of the junior roles. Data scientist is a distant third. Analytics engineer is a promotion, not an entry point.

A plan for someone already near data

Move the spreadsheet work you already do into SQL, ask for warehouse access, then the shared core: SQL until it is boring, one warehouse and one modeling tool, Python for data, a real project.

A plan from zero

Longer, and it works. A weekly group instead of a course, a nonprofit project for a portfolio piece and a reference, then the same core.

Your plan

Pick your starting point and honest hours per week. Tap each step for what to do and what to have at the end of it.

Start with where the jobs are

Two things stand out. First, only about one in sixteen responses is entry-level at all. This is a survey of people who visit a jobs board, so it skews experienced, but the shape matches what hiring managers say: junior roles are a small fraction of open positions and they get hundreds of applicants.

Second, the entry-level roles that exist are overwhelmingly two titles: data analyst and data engineer. Data scientist is a distant third. “Analytics engineer” barely has junior roles because the title is a promotion, not an entry point. If you are aiming at your first job, aim at analyst or engineer. Everything else is a second job.

of survey responses are entry-level

174 of 3,104

2 titles

hold most of the junior roles

analyst and engineer, roughly equal

4 in 10

entry-level roles are fully remote

about the same as everyone else

What AI actually changed

Not the thing people are worried about. Nobody I know has stopped hiring juniors because a model writes SQL. What changed is the definition of the job.

The old junior job was writing queries someone else specified. A model does that now, faster, and mostly correctly. The new junior job is knowing when the model is wrong. That means understanding the data well enough to notice that “revenue” came from the wrong table, that a join doubled the rows, that the trend is a timezone bug. Those are judgement skills, and they come from the same place they always did: doing the work on real data until the patterns are familiar.

So the plan above does not skip SQL. It makes you do more of it, by hand, until it is boring, because that is the only way to be able to check a machine that does it for you.

The plan, in prose

SQL first and longest. Six weeks minimum. Not a course, a dataset: pick one public dataset with a few related tables and ask it a hundred questions. Joins, window functions, CTEs. When you can write a query that answers “which customers ordered in three consecutive months” without looking anything up, move on.

One warehouse and one modeling tool. BigQuery’s free tier or DuckDB locally, and dbt or SQLMesh on top. This is where you learn the vocabulary every job posting uses: staging, marts, tests, lineage. A small project with a real DAG and a README is worth more than a certificate.

Python for data, not for software. Read a file, call an API, write a script that runs every day. pandas or polars. That is the whole syllabus. You are not becoming a backend developer.

One real project with a real question. This is the thing you will talk about in every interview, so pick a question that someone would pay to have answered. Public transit reliability, local housing prices, a nonprofit’s donor data. Publish a page with one chart that makes one point.

Then the step that depends on you. Students should get an internship or three merged pull requests to a tool they use. Career changers should rebuild the report their old team relied on, because knowing what the numbers mean is the rare skill. People in adjacent roles should move their existing spreadsheet work into SQL and ask for warehouse access, which most companies will grant. People starting from zero should find a weekly group and do a project for a small nonprofit, which gets you a portfolio piece and a reference at the same time.

Things that do not work anymore

Certificates alone. They were a weak signal in 2021 and they are noise now. Do the project instead.
A portfolio of Kaggle notebooks. Everyone has one and none of them answer a question anyone asked.
Applying to data scientist roles first. Look at the chart again. Aim at the doors that exist.
Waiting until you feel ready. The people who get in apply at about sixty percent readiness and learn the rest on the job. That has not changed.

What I am doing about it

I am putting together free, cohort-based sessions that follow the plan above, aimed at people who cannot pay for a bootcamp. If you want to be in the first group, or you want to help teach, subscribe below and reply to the first email. I will say more when there is more to say.