Data Science Interview Preparation

Compute the average salary by department and use a left outer join to list employee ids whose salaries exceed their department's average.

Compute each customer's highest and lowest total paid from a payments table by applying max and min window functions over partitions by customer id.

Identify customers whose status id equals 19 for three consecutive records, and print their customer id and origination date using a case statement and lead function.

Identify customers who never ordered by performing a left join between the customer reference table and the orders table, then filter where order id is null.

Learn to use the SQL lag function to compare each day's temperature with the previous day and identify dates or IDs where the current temperature exceeds the prior day.

Examine a dataset of user login details to determine each user's first login date by calculating the minimum login date per user ID using SQL.

Calculate monthly counts of new and repeated users by using first login as registration date, joining with login data, and aggregating with SQL to produce running totals.

Learn to calculate total logins per year, month, and day in SQL and derive a running sum with a window function.

Learn to identify the top three users per geography by total logins using a window function, partition by geography, and row_number filtering.

Practice SQL for interviews by writing a complete workable query first, explain your approach if needed, and train on LeetCode and HackerRank, focusing on joins, window functions, and aggregation.

Move zeros to the left and non-zeros to the right while preserving the non-zero order. Learn a Python notebook approach using two counters and end-to-begin iteration to implement the arrangement.

Demonstrates finding a pair of numbers in nums that sum to a target using nested for loops in Python, with an example showing 2 and 7 equaling 9.

Split a string on vowels in Python, replace vowels with a star, and use split to obtain the resulting substrings in a notebook.

Build a regression model to predict hourly bike sales using weather, temperature, humidity, and season, with data preprocessing, log transformation, and evaluation via XGBoost.

Develop a churn prediction model from customer data using xgboost, with preprocessing, log transforms, and one-hot encoding, then evaluate with log loss and a confusion matrix.

Learn to craft structured, interview-ready design answers by asking clarifying questions, defining goals and hypotheses, selecting data sources, and outlining models, metrics, and AB testing.

Design a system to predict infrastructure issues using clarifying questions, goals, and hypothesis-driven data. Build a production machine learning workflow from feature engineering to ab testing with clear metrics.

Design a real-time fraud identifying system by clarifying the problem, defining goals, detailing data sources and features, selecting models, evaluating with precision and recall, and planning A/B testing in production.

Learn how to test replacing the post composer with a bottom plus button, define goals and hypotheses, select metrics, and run A/B tests to boost engagement.

Define goals and hypotheses to evaluate product performance using metrics like new user acquisition, engagement, and feature usage, then test with randomized before-and-after A/B experiments.

Learn to identify and reduce fake accounts by building a machine learning model with logistic regression, using AB testing and risk-score metrics to protect user credibility.

Develop a data science interview strategy for charging an email product using clarifying questions, assumptions, and a/b tests to optimize open and click rates with machine-learning personalization from customer data.

Explore three probability problems: rolling a one then not six on two dice, two kids with at least one boy, and drawing two cards without replacement for color or suit.

Analyze dice probability: four ways to sum to five and eleven outcomes with at least one six, then compare mpg techs to show technology B saves more gasoline.

Apply hypothesis testing to assess if SAT scores must exceed 24, using X bar 24.8, mu 24, n=200, standard deviation 6.1, and reject null in favor of greater than 24.

Assess the sample proportion using a left-tailed hypothesis test to conclude that the failure rate is below 0.1 based on 500 phones (38 failures) with a p-value of 0.03.

Clarify precision and recall definitions, when to use each, and how F1 balances both, including false positives and false negatives, with notes on overfitting and regularization to improve generalization.

Examine bias, variance, and the bias-variance tradeoff, explore underfitting and overfitting, and cover linear regression assumptions, dimensionality reduction, and XGBoost versus random forest.

Explore multicollinearity and variance inflation factor to assess feature correlation. Learn p-values, null and alternative hypotheses, and logistic regression with sigmoid outputs to interpret coefficients and feature significance.

Learn to handle outliers without dropping them, perform univariate to multivariate analyses, compare R-squared and adjusted R-squared, and understand objective and loss functions, prediction probability, and log loss.