Question 1

Who is this collection for?

Accepted Answer

Candidates preparing for **Data Scientist, Senior Data Scientist, and Staff Data Scientist** interviews at top companies.

**Specifically:** Netflix, Stripe, Meta, Airbnb, Google, Amazon, Uber, DoorDash, LinkedIn, Microsoft, Apple, Salesforce, Spotify, Pinterest, Reddit — and 30+ other consumer-tech, fintech, SaaS, and marketplace companies.

This is the **complete DS interview prep path**; expect **~32 hours of focused work over 5-6 weeks**.

Works for **entry, mid, and senior** levels — the difficulty mix (12% d1, 29% d2, 43% d3, 16% d4) gives entry-level candidates ramp-up problems while still challenging senior candidates with d4 multi-concept capstones.

Question 2

How is this different from LDS Data Analyst Interview Prep 75?

Accepted Answer

**DS 75** and **DA 75** share LDS's question pool but answer fundamentally different interview prompts:

- **DA 75**: 7 topic-mastery stages, SQL-only, d2-heavy, no A/B testing depth, no stats inference, no ML
- **DS 75**: 6 interview rounds, **SQL + Python hybrid** (first cross-language LDS collection), d3-heavy, full A/B testing round (10 problems), inferential stats (8 problems), ML evaluation (3 problems)

If you're applying to **Data Analyst roles**, start with DA 75 instead. If you're applying to **Data Scientist roles** (or any role with "scientist" or "experimentation" in the title), DS 75 is the path.

Question 3

What's in each of the 6 rounds?

Accepted Answer

**Round 1 — SQL Screen** (10 problems, ~3h): Fundamentals — SELECT, WHERE, basic joins, aggregation. The 60-minute Hackerrank that gates every onsite.

**Round 2 — SQL Onsite Analytics & Time-Series** (12 problems, ~5h): Window functions, multi-CTE, time-series, ranking, running totals. The 90-minute onsite SQL round.

**Round 3 — A/B Testing & Experiment Analysis** (10 problems, ~5h): Variant comparison, lift, significance (Z-score, t-test, chi-square, ANOVA), sample-size analysis, CUPED. **The signature DS round.**

**Round 4 — pandas Coding Screen** (12 problems, ~4.5h): Filter, aggregate, merge, reshape, window, rolling. Python equivalents of Round 1-2 patterns.

**Round 5 — Statistical Reasoning & EDA** (14 problems, ~5h): Descriptive stats, IQR anomaly detection, correlation, hypothesis testing, bootstrap CI.

**Round 6 — Feature Engineering & Capstone** (17 problems, ~6.5h): Build features, train/test split, evaluate models. d4-heavy.

Question 4

What statistical and ML techniques does it cover?

Accepted Answer

Every technique used in real DS interviews per public reports (Interview Query, Glassdoor DS questions, Stripe/Meta/Netflix engineering blogs):

**Hypothesis Testing**: two-sample Welch's t-test (scipy.stats.ttest_ind), chi-square test of independence (scipy.stats.chi2_contingency), one-way ANOVA (scipy.stats.f_oneway), CUPED covariate adjustment.

**Inference**: bootstrap confidence intervals (numpy.random with fixed seed), Z-score for proportion-difference significance.

**Descriptive**: mean/median/std/quartile/IQR, Pearson correlation by group, outlier detection via 1.5×IQR rule.

**ML evaluation**: Linear regression (sklearn.linear_model.LinearRegression), R² and RMSE on train/test split, Precision/Recall/F1 + confusion matrix from raw predictions.

All powered by **pandas + numpy + scipy + sklearn** running fully in your browser via Pyodide — no install, no setup.

Question 5

Why 6 rounds, not 7 stages like DA 75?

Accepted Answer

**Because real DS interviews are round-based, not topic-based.**

When you onsite at Stripe/Netflix/Meta, you don't do a "subqueries round" or a "window functions round." You do:

1. A SQL screen (Round 1)
2. An onsite SQL round (Round 2)
3. **An A/B testing round** — the one that scares everyone (Round 3)
4. A pandas coding round (Round 4)
5. A stats round (Round 5)
6. A take-home or feature engineering case (Round 6)

The 6-round framing also gives DS 75 a different SEO surface than DA 75 — Google penalizes duplicate-template content, so we deliberately differentiated. Bonus: candidates Googling "what to expect in a DS onsite" land on a page that literally maps to those rounds.

Question 6

Do the Python problems run in the browser? What packages are supported?

Accepted Answer

**Yes — every Python problem runs entirely in your browser via Pyodide (Python compiled to WebAssembly).**

No install, no Conda, no virtualenv. Just hit "Run."

**Supported packages**: `pandas`, `numpy`, `scipy` (including `scipy.stats` for hypothesis tests), `sklearn` (including `sklearn.linear_model`, `sklearn.metrics`, `sklearn.model_selection`), `matplotlib`, and the Python standard library.

Not supported: `statsmodels`, `tensorflow`, `pytorch`, `xgboost` (yet — they're too large for the browser sandbox).

Question 7

Is this only premium / paid?

Accepted Answer

**Yes** — all 75 problems run on production-grade 8-15 table schemas modeled after real companies (*Stripe, Netflix, Meta, Salesforce, Zillow*, etc.). Schemas at that fidelity require **Premium** subscription access.

If you want **free** SQL prep first, try **LDS SQL 50** (free entry-level pool) before upgrading.

**Premium also includes:**
- AI Mentor (chat help during problems)
- step-by-step SQL Visualizer (see query execution per CTE)
- PostgreSQL/SQLite dual-dialect support
- Full A/B testing + statistics + ML evaluation reference solutions

Question 8

How long does this take?

Accepted Answer

**Estimated ~32 hours of focused practice**, typically **5-6 weeks** at 2-3 problems per day.

- **Round 3 (A/B Testing) and Round 6 (Feature Eng + Capstone)** are the densest — budget **30-40 minutes per problem**.
- **Round 1 (SQL Screen)** can be done at *10-15 min per problem* if you're comfortable with SQL basics.
- **Round 5 (Stats)** has the steepest learning curve for candidates without stats backgrounds — budget **20-25 min** even for the d2 problems.

Question 9

I have an interview next week. Can I still benefit?

Accepted Answer

**Yes — work backwards.**

If you have less than 2 weeks:

- **Round 3 (A/B Testing)** is the highest-leverage round to prep — interviewers love this.
- **Round 5 (Stats)** + **Round 6 (ML eval)** are the second-priority — they're the ones candidates fail on most.
- Skim Rounds 1, 2, 4 only on patterns you feel weak on.

For absolute crunch time, **Round 3 (A/B Testing) and Round 6 (ML Evaluation)** hold the most differentiated content — drill those first.

Question 10

Which SQL dialect does it use?

Accepted Answer

**SQLite by default** — runs in your browser via sql.js. **PostgreSQL mode** is also available via PGlite (also runs in browser).

The 27 SQL problems in this collection have **both reference solution (SQLite-native) and reference solution (PostgreSQL-native)** so you can practice in whichever dialect your target company uses.

All patterns taught transfer directly to **BigQuery, Snowflake, Redshift, MySQL, and SQL Server**. Date functions are the most-likely cross-dialect drift, and we use the most portable form throughout.

Question 11

How does this compare to DataLemur, StrataScratch, or LeetCode for Data Scientist interview prep?

Accepted Answer

All three are excellent at *parts* of DS prep. None of them covers the **whole** DS loop end-to-end.

**What LDS DS Interview Prep 75 gives you that they don't:**

- **A 6-round simulator covering all four interview surfaces** — SQL screen, SQL onsite, A/B testing, pandas, statistical reasoning, and ML evaluation
- **A dedicated A/B Testing round** (10 problems) — Z-score, Welch's t-test, chi-square, ANOVA, CUPED — the actual statistical methods DS interviewers use
- **A dedicated ML Evaluation capstone** — train/test split, confusion matrix, precision/recall, R²/RMSE — on real industry schemas
- **Full scipy + sklearn in the browser** via Pyodide — no install — so you actually run the t-tests and fit the models, you don't just write them
- **Industry-grade schemas** (Stripe-style, Netflix-style, Meta-style) — the kind of data DS interviewers ask about
- **Both SQLite and PostgreSQL** on every SQL problem

*DataLemur covers SQL well. StrataScratch covers breadth. LeetCode covers algorithms. None of them simulates the full DS onsite the way this collection does.*

LDS Data Scientist Interview Prep 75

About this collection

Your Learning Path

Why These 75?

All Four DS Interview Surfaces in One Path

A Dedicated A/B Testing Round

Real Statistics + ML in the Browser

Skill Coverage

FAQ

Ready to Master Data Science Interviews?

LDS Data Scientist Interview Prep 75

About this collection

Your Learning Path