At this level, everyone in the final round knows the fundamentals
Interview prep for mid-level through staff data scientists who are getting the interviews but not converting the final rounds. Built from 80+ interview rounds and 5 offers in Spring 2026, at Google, Uber, Meta, Figma, and Attentive. (Not getting interviews yet? Start with these free resources instead.)
In 2026: senior data scientist at Google, Uber, and Meta; staff at Figma and Attentive. I withdrew from four further final rounds.
Mid-level through senior staff, across more than 100 mock interviews. Neither list is exhaustive.
Getting the answer right isn’t what gets you the offer
At senior and staff level, everyone in the final round can code, knows applied ML, and can evaluate an experiment. Nobody’s getting the answer wrong. So not making mistakes stops being enough, because you’re up against people who also didn’t make mistakes.
The difference is nuance. In a case study, that means seeing who wins and who loses from a change, unintended consequences, and where the obvious metric choice can mislead you. In a project deep dive, it means leading with influence and impact instead of technical detail: what changed because you did the work, who you had to bring along, and where you got it wrong.
I went 1 for 3 in my first three finals this spring. Then I sat down, reviewed every single question and round I had faced, and rebuilt every answer expecting to get similar questions again. I took 4 offers from my next 5 final rounds before withdrawing from the rest. My technical ability didn’t change in those weeks. How I answered did.
The job changed this year, and the interviews changed with it
I’m an actively working senior data scientist who was personally going through extensive interviews this Spring. This matters more than it used to because the process has moved quickly over the past couple of years and anything written about it dates faster than it once did, including eventually this. The senior and staff offers I received came out of the interview formats companies are using right now rather than the topics common in 2021-2024.
The work itself has changed just as much: far less of my day goes into writing code and far more of it goes into directing a fleet of agents that can code exceptionally well but still make head-scratching analytical choices. That shift has worked its way into the interviews, where clean syntax carries less weight than it did a few years ago and product sense and hypothesis generation carry considerably more.
Case studies have broadened over the same stretch. They used to mean marketplace experimentation almost by default, but the same loop can now hand you an uplift modeling problem or consulting-style opportunity sizing. Candidates who only prepared for the first shape tend to get caught by the rest.
-
SpecializationExperimentation and observational causal inference — clustered and geo-based experiments, synthetic control, difference-in-differences
-
In industrySenior data scientist — Uber, Airbnb, Stripe. Currently in a full-time senior DS role.
-
Before techFederal Reserve and the SEC — research cited in a winning US Supreme Court case; expert economic witness testimony
-
TrainingPhD in economics, University of Michigan — peer-reviewed publications in economics and public policy
The Data Science Interview Playbook
In April and May 2026, I ran more than 80 interview rounds across sixteen companies and wrote up notes after every one. The Playbook is what came out of that, combined with lessons from over 100 mock interviews I have run with clients in the past two years.
It covers the full loop:
- The four case study archetypes — and how to tell within the first two minutes which one you are sitting in
- Coding rounds, including the AI-assisted format Meta introduced in May 2026 that most prep material predates
- Statistics and experimentation at the depth senior and staff candidates are actually held to
- Applied ML rounds — increasingly common, and came up at Google, Figma, Adobe, and Coupang this spring
- Past project deep dives, behavioral rounds, and the HR screen
- Practice problems closing every section, drawn from rounds I actually went through
The price reflects senior roles in US tech. If you are outside the US, or targeting roles where compensation is lower, tell me through the contact form and I will send you a code.
1:1 interview coaching
Mock case studies, guided walkthroughs, take-home reviews, and project deep dive prep, in a 45-minute session. Around 90% of what people ask me for is case study work, usually experimentation-related. I also take a shorter Ask Me Anything call for career questions, resume reviews, or a project at work you want a second opinion on, which is priced lower and is the wrong choice for interview prep.
Candidates often find themselves patching together fragmented blog posts, outdated practice problems, and scattered frameworks. I was hand-curating learning guides for my clients and was contemplating writing a comprehensive guide myself. Thankfully, Jonathan just solved this problem.
Yes, it’s a premium resource. But considering it distills the hard-fought patterns of 80+ actual interview rounds into an actionable blueprint, the ROI is a no-brainer. Land just one strong offer, and this pays for itself on day one.
Shared publicly on LinkedIn. Unpaid and unaffiliated.
A single session with Jonathan was worth more than 20 hours of studying for interviews. He walked me through interview structures and where to focus (and not to focus). Ultimately Jonathan helped me secure two offers, and the only thing I’d do differently is make sure I reached out sooner.
Working through the methods in public
I simulate data, run the models, and publish what actually happens, including when the result is inconvenient. Mostly causal inference and experimentation, occasionally interview strategy.
- The Rollout Trap: Why a Positive A/B Test Doesn’t Mean Treat Everyone Jul 2026
- The Waiting Tax: How a Surrogate Index Answers a 12-Month Question in 4 Jul 2026
- The Linear Trap: Why Double Machine Learning Beats Fixed Effects May 2026
- The False Positive Trap: The Dark Side of Bayesian A/B Testing Apr 2026
- Surviving the “Haircut”: Why Bayesian A/B Testing Beats Conservatism Apr 2026
Questions about any of it? Get in touch, and it comes straight to me.
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, like the AI-assisted coding rounds that showed up in May. When something shifts, I write it up and send it out.
No fixed schedule. I send when there’s something worth sending. Unsubscribe anytime.