DataLemur · LeetCode · HackerRank, with duplicates merged
0 / 0 done
Where you stand
0paid questions. Everything here opens free, so you never need an alternative or a subscription.
8free sources merged: LeetCode, DataLemur (free), HackerRank, Brainstellar, Project Euler, roadmap.sh and two GitHub answer banks.
10questions that appear more than once across these sources. Their boxes tick together, so you never solve one twice.
0items in this sheet after merging duplicates. 0 done.
What a data science interview looks like
Round
Length
What happens
Recruiter / HR screen
15–30 min
Background, why data science, notice period, pay.
Online test or AI interview
45–90 min
2–4 SQL queries, a pandas or Python task, and ML/stats questions: multiple choice, or spoken in an AI interview.
Take-home (some companies)
1–3 days
A dataset to clean, explore and model, plus a short write-up you present later.
Technical rounds (2–4)
45–60 min each
Live SQL or pandas, statistics and probability, ML theory, and a deep dive into your projects.
Product case / A/B test
45 min
Pick a success metric, explain why a metric dropped, design an experiment. Most common in product-analytics DS roles.
Hiring manager / behavioral
30–45 min
STAR stories: ownership, a failure, teamwork, learning fast.
The usual pattern in 2025–26 prep guides and candidate reports. Rounds vary by company, and startups often merge them into 2–3 calls.
What gets asked
SQL: joins, GROUP BY / HAVING, CASE, CTEs, window functions (RANK, LAG/LEAD, running totals), dates, top-N per group, percentages, retention and funnels.
Python: pandas (merge, groupby, pivot, dates, missing values) and easy-to-medium coding (strings, arrays, dictionaries). Sometimes a metric or a simple model in NumPy.
Statistics and probability: distributions, CLT, confidence intervals, p-values, power and sample size, Bayes, expected-value puzzles, A/B testing pitfalls.
Machine learning: bias–variance, regularization, metrics (precision/recall, ROC-AUC), trees, random forests, boosting, logistic regression, cross-validation, leakage, class imbalance, feature engineering, and "how would you build X?". LLM basics (embeddings, RAG, evaluation) now come up often too.
Product sense: metrics, diagnosing a drop, experiment design.
How the weighting changes by role
Role
SQL
Python
Stats & A/B
ML theory
Product case
DSA
ML system design
Data Scientist (product / analytics)
●●●
●●
●●●
●
●●●
–
–
Data Scientist (ML / modeling)
●●
●●
●●
●●●
●●
●
●●
ML Engineer
●
●●●
●
●●
–
●●●
●●●
Applied Scientist
●
●●
●●
●●●
●
●●
●●
Typical weighting; companies differ. Applied Scientist roles usually prefer an MS or PhD and go deep into ML math.
What this means for freshers
Fresher DS/ML roles, especially remote and startup ones, usually mean an online test or AI interview (SQL, pandas, ML/stats), then 1–2 calls that dig into your projects. Long product-case loops and DSA rounds are less common.
Coding (SQL, pandas) is where most freshers need the most practice, so most of the hours go there. For stats and ML theory, the work is explaining answers out loud in about 90 seconds.
If you're also preparing for GATE DA: it covers probability, statistics, linear algebra, ML and DBMS, so that prep counts as interview prep for those topics.
Heavy DSA practice mostly matters for MLE and Applied Scientist loops at large companies. It's parked in the "Later" tab.
Do you need a playlist like Striver's?
No. In data science the "sheet" is the problem list, which is this page. Learn each subject from one short source and spend the rest of the time on questions:
Topic
Learn from (one source)
SQL
DataLemur SQL tutorial (free, 30 lessons). It's in the plan: one lesson a day, with questions on what you've just learned.
pandas
Kaggle Learn: Pandas (free, 6 lessons, about an hour each with the exercises: one lesson a day).
Udacity: A/B Testing by Google (free), lessons 1, 3, 4 and 5 only. The plan slots them in among the A/B prompts in the Stats tab.
Order of attack: every subject, every week
Don't finish SQL before starting the rest. Each week you do some SQL, pandas, stats and probability, and ML, plus re-solves of older questions, so nothing goes cold. The Today's plan tab picks the exact questions for each day and brings old ones back for review. It starts the first day you open it.
Monday to Saturday, about 3 hours (more on days a topic is new): 2 SQL items (a tutorial lesson plus a question on it, until the tutorial is done) and 1 pandas item every day, then either 2 stats and probability questions (Mon, Wed, Fri) or 3 ML questions answered out loud (Tue, Thu, Sat), then the re-solves due that day, mostly the questions you did a week earlier.
Sunday: a light day. 1 project or A/B item, plus any re-solves due or left over from the week.
At this pace the 500 core items take about 20 weeks.
After that: polish. Project deep-dives, a practice take-home, mock interviews.
Why there are no paid questions
Paid questions on DataLemur and LeetCode are left out. The free problems that train the same patterns are already in the lists; in SQL D those are marked core instead of optional.
DataLemur's paid stats and ML questions are mostly concept questions. The free GitHub answer banks and Brainstellar cover the same concepts with answers on the page.
A question that appears on two sites in the same words is listed on both but ticks together, so you see it as done wherever you meet it.
How to use this sheet
Tick a box when you've solved it. Ticks are saved in this browser, so clearing its history and site data, or switching browsers, resets them.
Items that are the same problem on two sites tick together, so you never solve one twice.
For GitHub answer-bank items, the link opens the question (or its topic for Grigorev's list) with the answer underneath. Answer out loud before you scroll down.
Search also matches concepts (try "window" or "A/B"). The level filter, "Core only" and search narrow the list, and every count follows them: the tab counts, each group's count and the line under the filters.
Core only hides the 140 items marked optional and leaves the 500 core items. The optional ones are HackerRank's Easy warm-ups and its SQL (Advanced) certificate test, extra LeetCode SQL practice, 2 "deadly" puzzles, the deep-learning, text, ranking, recommender and time-series theory, and the "Later" tab.
Dates assume you follow the weekly plan below. When a subject's core list is finished, its daily slots move to its partner (SQL and pandas swap, stats and ML swap), so the rest speeds up.
The weekly plan
About 3 hours a day, more while a topic is new. SQL and pandas get a slot every day because coding is where most people need the most practice. Theory alternates between stats and ML.
Day
SQL
pandas / Python
Theory
Review
Mon, Wed, Fri
2 new
1 new (or 1 Kaggle lesson)
2 stats & probability
last week's questions, once each
Tue, Thu, Sat
2 new
1 new (or 1 Kaggle lesson)
3 ML questions, out loud
last week's questions, once each
Sunday
No new questions. 1 project or A/B item from "Case & projects".
anything left over
Per week: 12 SQL, 6 pandas, 6 stats and probability, 9 ML and 1 project or case item, plus one re-solve of each question a week later (about 33, roughly 30–45 minutes a day).
How the review stops you forgetting
When you tick a question, it comes back here once, 7 days later, for a re-solve from memory. Press Re-solved and it's done.
Re-solve from a blank editor, or answer out loud. If you couldn't do it, leave it: it comes back the next day until you can.
Every re-solve that's due is on the list, so nothing piles up or comes back late. Courses and project or A/B items don't come back.
A new day starts at midnight (12 am).
Rules that make it work
The order inside each subject is set for you:
SQL: alternates LeetCode SQL 50 and DataLemur, then HackerRank Medium and Hard, then the core items of D.
pandas: starts with the Kaggle course, one lesson a day (6 lessons, about an hour each), then LeetCode's two pandas plans, then DataLemur Python.
Theory: mixes puzzles and concept questions.
Stuck for 20 minutes? Read the solution, write it yourself, and leave the box unticked. It comes back the next time that subject is on the plan.
On a busy day, do just the SQL and the review.
Today's list stays fixed for the day. Ticking a question anywhere in the sheet also ticks it here.
One resource per subject
Everything you need is already in the tabs and in Today's plan. These are the only places you learn from; the rest is practice.
Kaggle Learn: Pandas (free, 6 lessons). Practice comes from LeetCode's two pandas study plans.
Statistics and probability
Any one probability and statistics course you already use; Khan Academy statistics if you have none. Practice comes from DataLemur, Brainstellar and the GitHub Q&A.
Machine learning
StatQuest on YouTube (free). Practice comes from DataLemur, Grigorev's and youssefHosni's answer banks and roadmap.sh.
A/B testing
Udacity: A/B Testing by Google (free), lessons 1, 3, 4 and 5 only, then the 10 A/B prompts in the Stats tab. The Cookie Cats project in Case & projects is optional.