Data Science Career Guide - Interview Preparation

Discover why data science offers fulfilling, high-demand work across industries by applying math and coding to extract insights, with strong pay and growing job opportunities.

Explore how data science blends computer science, math and statistics, and domain knowledge into an interdisciplinary practice, and learn how T-shaped candidates combine breadth with deep domain expertise.

Explore the spectrum of data science roles beyond data scientist, including product analyst, business intelligence engineer, machine learning engineer, and data engineer, and how job postings define required skills.

Examine how graduate degrees influence data science careers, from large company roles to freelancing, using Ian Goodfellow and Adam Gibson as extremes to show the value of portfolios.

Explore a high level overview of what you should know for a data science career, covering technical tools, software and languages, probability and statistics, machine learning concepts, and job-specific tools.

Delve into programming languages, frameworks, and software tools in data science, and learn to pick one language (often Python) plus SQL and key libraries like pandas and scikit-learn.

Develop a solid theory backbone for data science, an interdisciplinary field, covering calculus, linear algebra, probability, statistics, Bayesian statistics, AB testing, and descriptive statistics, with on-the-job learning.

Explore core machine learning concepts for data science roles, including supervised and unsupervised learning, key algorithms, model validation, regularization, and basics of natural language processing, with practical resources.

Learn software engineering knowledge for data science interviews, including algorithms, data structures, SQL fundamentals, and the potential use of Spark, Hadoop, and data visualization tools like Tableau or Plotly.

Assess readiness for data science roles by demonstrating working knowledge of programming, knowledge of probability and statistics, and general knowledge of machine learning, using practice interview questions from real recruiters.

Learn to craft data science resumes with quantified achievements, customize for top target roles, and build social proof via LinkedIn, GitHub, Kaggle, and more, all while keeping to one page.

Navigate the data science interview pipeline—from connection and recruiter screening to technical screening, take-home tasks, and on-site interviews—highlighting how automation and company variations shape the process.

Navigate the data science interview funnel for your first position, from applying to 100–200 jobs to phone, technical, take-home, and onsite screens, using networking and projects to boost success.

Navigate offers by evaluating total compensation, including salary, bonuses, equity, and benefits, and compare industry data to negotiate up to 10–15% more while discussing vacation, relocation, and review schedules.

Explore common probability theory interview questions and practice with coins and dice to prepare for data science technical screenings.

Explore the expected number of flips required for a fair coin to yield two consecutive identical results, whether two heads or two tails, in this probability theory interview question.

For a fair coin, the lecture solves the probability of obtaining two consecutive flips of the same result and shows the expected number of flips is three.

Explore the probability of obtaining a total of four when rolling two dice by adding the outcomes and calculating the likelihood.

Compute the probability of rolling a sum of four with two dice by counting three favorable outcomes—(1,3), (2,2), (3,1)—out of 36 total outcomes, giving 1/12.

Explore the probability of rolling at least one four with two dice by focusing on the dice numbers, not their sums, in this probability theory interview question.

Explore the probability of rolling at least one four with two dice by counting 11 successful outcomes out of 36. The lecture shows the method and result of 11/36.

Explore a probability puzzle: distribute 50 red and 50 blue marbles into two jars to maximize the chance of drawing a red marble when first choosing a jar at random.

Place one red marble in jar one and all remaining marbles in jar two to maximize the chance of drawing red, yielding about 74.75% probability.

Learn to compute the probability of seeing a car in ten minutes when the chance in thirty minutes is 0.95, assuming a constant probability.

Assume a constant default probability to solve the interview question. Compute P from 1−P^3=0.95, then the ten-minute car probability is 1−P.

Compute the average number of flips needed to obtain two heads in a row with a fair coin, an interview question focusing specifically on two heads rather than tails.

Learn how to compute the expected number of flips to obtain two heads in a row with a fair coin, arriving at an expected six flips and a case analysis.

Compute the probability the drawn coin is biased given three flips result in heads, heads, tails, with nine fair and one biased coin whose heads probability exceeds 0.5.

Apply Bayes' theorem to determine the probability the chosen coin is biased after observing heads, heads, tails from a mix of one biased and nine fair coins.

Explore how to simulate a fair coin using a biased coin with heads probability greater than 0.5, a tricky probability theory interview question often asked by Google.

Discover how to simulate a fair coin from a biased one with two independent flips: heads-tails equals heads, tails-heads equals tails, discarding same-face results.

Solve a probability theory interview question about two children, where one is a girl, and determine the probability the other is also a girl, assuming equal odds for each gender.

Explain a two-child probability problem by conditioning on at least one girl, narrowing to three outcomes, and showing a one-third chance of a girl and two-thirds of a boy.

Prepare for statistics interview questions by applying probability concepts and distributions such as normal and binomial, using the resource guidebook and table lookups as you would in take-home interviews.

Assess the probability that it is raining in Seattle based on three independent friends who may lie or tell the truth, starting from a 25 percent prior.

Use a probability tree to determine the likelihood it is raining in Seattle given three independent reports. Leverage truth and lie probabilities to compute the conditional rain chance.

Solve a statistics interview problem by calculating the probability of seven successes in ten attempts with p = 0.8 using the binomial distribution.

Explore the binomial distribution with a 0.8 success probability to compute seven successes in ten attempts, using the binomial formula and independence of trials.

Explore the difference between a type one error and a type two error in statistics, and practice explaining it clearly by whiteboarding your answer.

Explore type one and type two errors in hypothesis testing—false positives and false negatives—when the null hypothesis is true but rejected, or false yet not rejected, with toothpaste examples.

Apply Bayes' theorem to a medical test: with 1% base rate, 99% sensitivity, and 99% specificity, a positive result indicates a 50% chance of infection.

Apply Bayes rule to medical test results to determine illness probability after a positive test, showing how low base rates yield about 0.5 posttest probability despite high accuracy.

Explore how to determine a motor guarantee using a normal distribution with mean 10 years and standard deviation 2 years, targeting the 3% failure tail via a z-table.

Use the z-table to map the 3% tail to a z-score of -1.88 in a normal distribution with mean 10 and standard deviation 2, then solve for x to set a 6.24-year motor guarantee.

Learn to tackle open-ended product design and metrics questions in data science interviews by thinking aloud, framing answers as a collaborative conversation, and analyzing features tied to company metrics.

Develop a structured approach to product design and metrics for evaluating a standalone messenger app, focusing on performance tracking and metric selection in an interview context.

Analyze metrics for launching a standalone Messenger app, focusing on daily active users, monthly active users, total time spent, messages sent, and response times to assess engagement and advertising impact.

Analyze an ab test where a flawed new search algorithm yields less relevant results, causing the treatment group to perform more queries and earn higher ad revenue.

Analyze a product design and metrics interview scenario about an ab test where a new search algorithm gives less relevant results, increases queries, triggers clarifying searches, and affects ad revenue.

Explore how A/B testing with a new search algorithm can yield higher advertising revenue despite less relevant results, and examine potential causes within product design and metrics.

Explain why a new search algorithm can raise advertising revenue despite less relevant results by increasing searches and potentially more relevant ads served by a separate ads algorithm.

Evaluate whether to implement a less relevant Google search algorithm when AB tests show higher advertising revenue and more queries in the treatment group.

Evaluate an AB test where a less relevant algorithm increases short-term ad revenue, and conclude not to implement it to protect long-term site relevance and growth.

Evaluate product design and metrics by comparing mpg upgrades for Technology A on car X and Technology B on car Y, with a 50/50 country split, to maximize gasoline savings.

Analyze two fuel-efficiency policies in a product design and metrics interview, showing policy B saves more gasoline countrywide by comparing mpg improvements with an average distance D.

Master basic SQL concepts and common interview questions to demonstrate data access skills. Learn to recognize SQL flavors and engine syntax, and practice writing queries from tables.

Identify what's wrong with a sample sql query, pause to review. Prepare to learn the solution in the next lecture.

Learn why you should not use the trial year alias in a where clause, and filter with year(trial_date) less than or equal to 2015 to correctly apply the condition.

Identify the error in the SQL query select ID trial date from payments group by ID, as presented in the data with SQL interview question that highlights grouping and selection issues.

Learn how to fix a SQL group by error by applying an aggregate to non-grouped columns like trial date. Group by date when dealing with time-stamped values to clarify results.

Analyze a flawed sql query that selects user ID and avg(total) as average order total from invoices, using a having clause with count on order ID >= 1.

Spot the missing group by in a SQL query that uses avg on the total column to compute the average order total from invoices, using having.

Write an SQL query to join the employees and managers tables on the managed_by foreign key. Retrieve all employees who are managed by Sandy Kim.

Master SQL joins to solve an interview question: find employees managed by Sandy Kim. Build a join on employees.managed_by and managers.id, then filter where managers.name like 'Sandy Kim'.

Write and practice a query that retrieves all employees with no manager using the employees and managers tables, illustrated by Jane Doe's null manager.

Retrieve all employees who have no manager using sql query that checks for null in the managed by column without join, noting variations across mysql, postgresql, oracle, and sql server.

Explore theory-focused machine learning interview questions on linear and logistic regression, decision trees and random forests, naive Bayes, SVMs, model evaluation and training, with whiteboard reasoning.

Explore the main assumptions and common types of linear regression, and review the most widely used estimation techniques for machine learning interview questions.

Explains linear regression and its core assumptions—linearity between y and x and normally distributed residuals—and reviews common types like ordinary least squares and ridge and lasso.

Describe the logistic regression formula and how it enables binary classification within the machine learning interview questions, as part of the data science career guide.

Describe the logistic regression formula and how to use the logistic function to turn linear regression outputs into probabilities for binary classification, with a 0.5 cutoff.

Explore how a decision tree chooses splits by examining the criteria for split points in machine learning interview questions.

Explore how decision trees choose splits by maximizing information gain using entropy in a top-down approach from the root node, with alternatives like the Gini index.

Explore the advantages of a decision tree model over other machine learning models, and prepare for the solutions discussed in this lecture.

Provide easy interpretation of decision trees, handle both continuous and categorical features, require no normalization or scaling, and deliver fast predictions on large data sets.

Explain the difference between random forest and boosting tree algorithms such as gradient boosting, highlighting their approaches and use cases for machine learning interview preparation.

Contrast boosting trees with random forests by noting that boosting reweights hard-to-classify samples iteratively, while random forests rely on bootstrap aggregation and random splits to build multiple trees.

Explain the assumptions naive Bayes methods make for a data set of features X and labels Y.

Identify the Naive Bayes assumption that X features are conditionally independent given Y. Note this independence may not hold, but we apply the model under this simplifying assumption.

Describe how the support vector machine works in a general sense and illustrate the concept with a diagram to explain decision boundaries.

Explore how support vector machines find a hyperplane that maximizes the margin between classes, using support vectors to define the decision boundary, and how the kernel trick enables nonlinear classification.

Define overfitting in machine learning, discuss its causes, and outline ways to avoid it in practice.

Learn how overfitting arises when a model fits training data too closely and fails on unseen data; explore causes and prevention techniques like regularization, early stopping, and cross-validation.

Describe the differences between accuracy, precision, and recall in classification tasks. Understand these common metrics and how their definitions relate to model performance.

Explore accuracy, precision, and recall in binary classification, using true positives, true negatives, false positives, and false negatives, and learn their formal formulas.

Identify metrics for evaluating a regression task and prepare for a machine learning interview with practice questions and solutions.

Learn to evaluate regression with mean absolute error, mean squared error, and root mean squared error. MAE averages absolute errors, MSE squares errors, RMSE preserves units.

Explore the design of experiments and its statistical foundations, including P value, statistical test, and hypothesis test. Review concepts before attempting the questions and consult the guidebook for resources.

Explore design of experiments for testing user reception of a new button on Google's main search page, outlining processes to determine whether users enjoy the feature.

Master design of experiments with a/b testing by selecting metrics like daily active users, setting up control and variant pages, randomizing samples, and evaluating hypotheses using alpha and p-values.

Learn how to detect sample bias in design of experiments and understand the different biases to watch for when evaluating data samples.

Explore how to detect sample bias in design of experiments by bootstrapping sample means around the population mean, and examine selection and measurement biases with practical examples.

Apply design of experiments to test up to 50 homepage variations, define metrics, and report results with a structured setup beyond simple A/B testing.

Explore designing experiments with 50 homepage variations, performing t-tests among pairs, and applying Bonferroni correction to adjust alpha for multiple hypotheses, especially with large user populations.

Explore the definition of power for a statistical test, the factors that affect it, and how power relates to the P value in design of experiments.