The data science resume template
The structure I use on my own resume and LinkedIn profile, with the reasoning behind each section. Free, and on this page rather than behind a download.
I rarely use referrals, and I still get recruiter callbacks from cold online applications at companies that reject most of what they receive. Your background does most of that work, and I am not going to pretend otherwise. But plenty of candidates have backgrounds that read as similar on paper, and among those, presentation changes the response rate more than people expect.
You need two things to get interviews from cold applications: relevant experience, and a presentation that makes it legible in about eight seconds. This page handles the second one.
What the layout looks like
This is the structure of my own resume with the specifics stripped out. Fill in your own names, dates, and numbers. The bracketed placeholders mark where a detail goes; the shape of the bullets is the part worth copying.
- Designed and analyzed geo- and cluster-randomized experiments for decisions where user-level randomization was invalid, covering the power analysis, the readout, and the launch recommendation.
- Rebuilt the team’s measurement approach for [channel or product area] after showing the existing specification misattributed effects across the distribution, changing how a $XXM budget was allocated.
- Estimated the incremental effect of [initiative] against a holdout, and found platform-reported performance overstated it by roughly XX%.
- Served as the reviewer for experiment designs across the team, and presented causal results to [senior stakeholders] during budget reviews.
- Built synthetic control frameworks to evaluate initiatives that could not be randomized, including repricing legacy free products, which preserved $X.XM in monthly revenue.
- Measured the cost of [a negative user experience] using propensity scoring and inverse probability weighting, where running an A/B test would not have been ethical.
- Designed A/B tests in the live checkout flow to estimate price elasticity, which let the business pass through a vendor surcharge without losing volume, saving over $XM.
- Owned a machine learning [lead scoring or targeting] model optimizing for expected value, validated with a clustered experiment that surfaced an upstream data error before it reached production.
- Collected and standardized panel data on [subject] from [public data source].
- Modeled expected outcomes from pre-period data, then identified the predictors of the gap between actual and predicted results.
- Ran fundraising, increasing annual grants from $10,000 to $25,000.
How to write the career bullets
Career experience is where I see the most damage, and it is almost always the same problem: the bullet describes an activity rather than a result. Write these the way you would answer in an interview, following situation, task, action, result.
Here is a line I have read on many resumes, and the same work written properly.
Responsible for communicating with multiple stakeholders.
A reader in a hurry cannot tell whether this means explaining causal estimates to a department head or sitting on calls reading out descriptive statistics. Given the choice, they assume the second.
Designed and analyzed experiments to estimate the impact of repricing legacy free products, driving $2.5 million incremental revenue per year.
The situation is products that were free and might not stay that way. The task was sizing the change, the action was designing and running the experiment, and the result is a number.
What belongs where
-
Order of sections
Experience goes above technical skills once you have relevant experience, because that is what a recruiter is scanning for. Reverse it if you are early in your career and your projects and coursework are carrying the page. Projects, publications, and hobbies are optional sections you add only when the required three do not make the case on their own.
-
Header
Name, email, phone. Use a Gmail address or a business one, close to your actual name, with no nicknames in it. An old Yahoo or Hotmail address will not sink you, but it does make interviewers wince.
-
Education
Recent graduates with high grades should include a GPA on a four-point scale. I would avoid anything above 4.0 in the numerator, because it is imprecise and it makes a reader wonder what the real scale was. If space runs short, drop degrees a later one supersedes.
-
Technical skills
Split these by track rather than dumping every tool into one line, so a reader looking for a specific skill set finds it in one place. SQL is by far the most common technical skill in data science interviews, and it should appear in more than one spot on the page. Nobody expects Spark or Airflow from an entry-level analyst.
-
Professional experience
This is where I see the most damage. You are trying to show three things at once: technical skill, business judgment, and impact. Avoid anything that could be read as not requiring a technical background, because a reader in a hurry will read it that way.
-
Projects
There is a wide gap between classroom and practical work, and projects are how you close it when your career experience does not yet cover the roles you want. If you are choosing a capstone, expect it to become the bulk of your interviews, so pick one you will enjoy defending.
-
Presentations and publications
Mostly relevant for PhDs, but worth listing even when the subject is far from data science. Presenting or publishing shows you took something to the level of expertise, which is a signal that survives a change of field.
-
Hobbies, activities, leadership
Useful mainly for recent undergraduates with space to fill. A serious hobby or a real leadership role gives an interviewer something to connect over. Expect this to matter less at large companies, where interviews are standardized to strip out exactly that kind of personal read.
-
Rebuilding it in Word
The alignment comes from tables rather than tab stops: company, school, and job title in the left cell, dates and location in the right, with the borders set to white at 0pt so they disappear. Build it with visible borders first and hide them at the end. I use 13 point for the section headers and 11 for everything else, and I would not move more than one size in either direction.
When a second page is fine
One page is the standard and mine has been two for years. A second page is fine as long as everything on it is additive. Older work that is junior, or a repeat of what you do now, can sit in a line or two without bullets, and should never be the reason a resume runs long.
The cases that earn a second page are the ones where your last two roles are both relevant and the earlier work adds something different, such as consulting or marketing. Conference presentations and publications earn it too, which comes up often for PhDs moving into data science.
What has changed about applying online
Applying online has gotten harder every year, and 2026 is the hardest version of it I have seen. A large share of candidates now write their resumes with AI, so most of what a recruiter opens is clean, conventionally formatted, and free of obvious red flags.
That is not a complaint, because for technical roles a standard format is probably the right strategy, and it is what this template gives you. But it does mean formatting is no longer where you win, because almost everyone else has cleared that bar too.
Write about categories of problems, not one company’s version of them
Recruiters are looking for evidence that you have already done the thing they need done, which is why specificity helps. The risk is describing it so specifically that you read as someone who can only do it at your last company, on their data, with their tooling.
The specificity cuts both ways, because it weakens the signal for anyone whose stack looks different and invites the assumption that you would be hired to run the same process again. Write for the category of problem, and let your company be one example of it.
Designed experiments to optimize the CHOICE v1 recommender model against the GALAXY v2 model.
Nobody outside that company knows what CHOICE and GALAXY are, so this reads as internal maintenance rather than a transferable skill.
Partnered with ML engineers to design and analyze experiments testing competing recommendation models. Owned the power analysis, metric choice, and readout, leading to an X% improvement in completion rates.
The same work, legible to any company running experiments on a ranking system, and clear about which parts you personally owned.
The same applies in interviews when you talk through past work. If your project deep dive relies on internal names and internal context, the interviewer has to do the translation for you, and most will not bother.
Knowing when to stop editing
Marginal changes to a resume are almost impossible to evaluate. You rarely learn why an application went nowhere, and in a pile this size the difference between a good resume and a slightly better one may genuinely be nothing.
So if yours follows a standard format, explains your work clearly, and you are still not getting interviews, the resume is probably not your constraint. That is the point to move your effort to networking, where one warm introduction does more than another round of edits. The networking guide is here, and it is also free.
What interviewers are actually asking, as I hear it
I coach people currently interviewing at OpenAI, DeepMind, and the big tech DS teams, so I see format changes early. When something shifts, I write it up and send it out.
No fixed schedule. I send when there’s something worth sending. Unsubscribe anytime.
Once the resume is getting you interviews
Getting into the room and performing once you are there are different problems. The Playbook covers the second one: the case study, the coding round, the project deep dive, and the behavioral questions, at the senior and staff bar.