If you want to pursue a career in data science, you must train thoroughly and leave a lasting impression on prospective employers with your experience in the field. 

During a data science interview, an interviewer will ask various kinds of data science interview questions, encouraging the interviewee to demonstrate robust scientific expertise and strong communication skills. 

The interview will put your knowledge of statistics, programming and data modeling to the test. A series of questions and problem types will challenge you to demonstrate your ability to perform under pressure. This article provides a comprehensive list of data science interview questions that you can prepare for before applying for a data scientist job.

List of Data Science Questions

  1. Fundamental Interview Questions On Data Science 
  2. Data Analysis Interview Questions For Data Science 
  3. Machine Learning Interview Questions For Data Science
  4. Python Interview Questions For Data Science
  5. Deep Learning Interview Questions For Data Science
  6. Statistics Interview Questions For Data Science 
  7. R Interview Questions
  8. TensorFlow Interview Questions
  9. Mahout Interview Questions And Answers

1) Fundamental interview questions on data science 

Below are some of the most basic data science interview questions and answers.

  1. What is selection bias? 

Selection bias is a type of error that happens when a researcher selects the study subjects. It is typically correlated with experiments in which subjects are not randomly chosen. This phenomenon is sometimes referred to as the selection effect. It is an illusion in statistical analysis caused by the sample collection process. If selection bias is not considered, a few of the study’s results might be incorrect. There are several types of selection bias:

  • Bias in sampling: It is a statistical mistake caused by a non-random selection of a group. Some participants are less likely to be included than others, resulting in a skewed sample. 
  • Interval of time: Since a trial can be stopped early if an extreme value is achieved (often for legal reasons), the extreme value is more likely to be reached by the variable with the greatest variation, even if all variables have a similar mean. 
  • Data: Data bias happens when subsets of data are selected arbitrarily to justify a hypothesis or dismiss bad data, rather than according to previously stated or universally accepted parameters.
  • Attrition: Attrition bias is a form of selection bias triggered by attrition (participant loss), resulting in trial subjects/tests not being performed.
  1. What is the difference between confidence intervals and point estimates? 

Point estimation provides an approximation of a population parameter using a given value. The System of Moments and Maximum Probability estimators are used to obtain point estimators for population parameters. 

A confidence interval indicates the number of possible values for a population parameter. The confidence interval is typically chosen as it suggests the probability of this interval containing the population parameter. This likelihood of likeliness is referred to as the confidence level or confidence coefficient and is denoted by 1 — alpha, where alpha is the significance level. 

  1. What is A/B testing’s purpose? 

It is a check of hypotheses for a randomized trial involving two variables, A and B. 

A/B testing aims to detect any improvements that may be made to a web page to optimize or improve the desired outcome. A/B checking is an excellent tool for determining the business’ most effective web promotion and marketing tactics. It can be used to test almost everything, from website copy to sales emails to search advertisements.

Basic Python interview questions for Data Science are the most commonly asked questions. Make sure you have the basics clear.

  1. What is a p-value? 

A p-value will help you evaluate the accuracy of the findings while running a hypothesis test in statistics. The p-value is a positive integer between 0 and 1. The value indicates the strength of the analysis. The argument that is being tested is referred to as the Null Hypothesis. 

A small p-value (0.05) implies strength against the null hypothesis, implying that we should reject it. A high p-value (0.05) means that the null result is strong, suggesting that we can consider it. A p-value of 0.05 indicates that the conclusion could go in any direction. High p-values indicate that the data contains a real 0. Low p-values indicate that the data are unlikely to have a true 0. 

  1. Distinguish between univariate, bivariate and multivariate statistical analysis. 

There are descriptive mathematical research methods that can be classified according to the number of variables present at a specified time point. For instance, pie charts representing revenue by region include only one variable and are therefore referred to as univariate analysis. 

An investigation is conducted to determine the variance of two factors simultaneously, such as in a scatterplot. It is referred to as bivariate analysis. For instance, the bivariate analysis may be used to examine the amount of revenue and expenditures. 

Multivariate regression is used to describe more than two variables to ascertain the influence of the variables on the responses.

  1. How are supervised and unsupervised learning different? 

When an algorithm learns from testing data to apply the information on test data, this is referred to as supervised learning. Classification is an illustration of supervised learning in action. Unsupervised learning happens where the algorithm does not learn something beforehand due to the lack of an answer variable or training details. Clustering is an unsupervised learning technique.

  1. What is the difference between an Eigenvalue and an Eigenvector? 

Eigenvectors help in the comprehension of linear transformations. In data analysis, the eigenvectors of a correlation or covariance matrix are commonly measured. Eigenvectors are about how a linear transition flips, compresses or extends. The eigenvalue is considered the power of transformation in the eigenvector’s path or the element by which compression happens. 

  1. Do gradient descent approaches converge on the same point invariably? 

They do not, like in certain situations, hit a local minima or maxima value. You may not arrive at the global optima point, and the data and initial circumstances determine it. 

  1. Explain the box cox transformation and how it is used in regression models. 

For any cause, the response variable in a regression analysis could violate one or more of the standard least squares regression assumptions. The residuals can adopt a normal distribution or a distorted distribution as the projection rises. In such cases, the answer variable must be transformed to ensure that the data satisfies the requisite assumptions. A box cox transformation is a mathematical method used to normalize non-normal dependent variables. If the data is not usually distributed, the bulk of mathematical methods can presume normality. By using a box cox transformation, you will conduct a more excellent range of experiments. 

You can also be asked some Data Scientist interview questions, so be prepared for that too!

  1. Which data scientists do you admire the most?

There are too many excellent startups in data science, and naming them all can be an arduous task. Here are a few big names you can cite as your favorite data scientist. 

  • Geoff Hinton, Yann LeCun and Yoshua Bengio
  • Demis Hassabis
  • Jake Porway from DataKind
  • DJ Patil
  • Kirk D. Borne
  • Claudia Perlich
  • Hilary Mason
  • Usama Fayyad
  • Hadley Wickham

2) Data analysis interview questions for data science 

Data analysis is the method of transforming data to uncover relevant facts that can be used to draw conclusions or make decisions. Data analysis is commonly employed for a variety of applications in all industries. As a result, there is a high demand for data analysts on a global scale. Top data science interview questions for freshers in the data analysis domain are listed below.

  1. Explain data cleansing. 

Data cleaning, alternatively referred to as data cleansing, deals with finding and deleting anomalies and discrepancies from data to increase its consistency. 

  1. Explain the differentiation between data mining and data profiling. 

Data profiling is concerned with the study of individual attributes on an individual basis. It contains knowledge about various properties, including value ranges, distinct values and their frequency of presence, the occurrence of null values, data sort and weight. 

Data mining is concerned with cluster processing, the identification of irregular records, the discovery of dependencies, the discovery of sequences and the creation of relationships between multiple attributes. 

  1. What is the KNN imputation method? 

In KNN imputation, missing attribute values are imputed using the values of the more identical attributes to the absent attribute. The similarity of the two attributes is calculated using a distance function. 

  1. What is collaborative filtering and how does it work? 

Collaborative filtering is a straightforward algorithm for developing a suggestion method centered on user experience. Collaborative filtering’s most critical elements are users-items-interest. Collaborative filtering is shown when a statement such as “recommended for you” appears on online shopping pages, depending on your browsing history. 

  1. What is correlogram research and how does it work? 

In geography, a correlogram analysis is a general form of spatial analysis. It is a set of approximate autocorrelation coefficients for various spatial relationships. When the raw data is represented as distance rather than values at individual points, it can be used to create a correlogram.

  1. What is an affinity diagram? 

An affinity diagram is a type of analytical diagram used to group or arrange data based on their relationships. These data or ideas are generated primarily by discussions or brainstorming sessions and it analyzes complicated problems. 

  1. What is data visualization and how does it work? 

In simple words, data visualization is the depiction of knowledge and data graphically. It allows users to interpret and evaluate data more intelligently and visualize them using graphs and charts created with technology. 

  1. What exactly is Metadata? 

Metadata is a word that relates to the basic information about a data structure and its contents. It assists in determining the kind of data or details to be sorted. 

  1. What is MapReduce? 

MapReduce is a method of developing applications that handle vast data sets by partitioning them into subsets, processing each subset on a separate server, and then combining the findings produced on each server. It is composed of two tasks: Map and reduce. The map function filters and sorts, while the reduce function executes a summary process. As the name implies, the reduce operation is done after the map operation. 

  1. What is SAS interleaving? 

Interleaving in SAS refers to merging several sorted SAS data sets into a single sorted data set. By combining a SET declaration and a BY statement, data sets may be interleaved.

3) Machine learning interview questions for data science

Machine Learning is an essential part of data science. Some of the top data science interview questions in the domain of machine learning (ML) are listed below.

  1. How can you define the terms, supervised and unsupervised ML? 

We must have classified data for supervised machine learning algorithms, such as stock market price prediction, but not for unsupervised machine learning algorithms, such as email sorting into spam and not spam. 

  1. Distinguish between KNN and k.means clustering. 

K-Nearest Neighbors is a supervised ML algorithm in which we supply the model with labeled data, and it classifies the points depending on their distance from the nearest points. K-Means clustering is an unsupervised ML method, we must provide unlabeled data to the model. This algorithm categorises points into clusters based on the mean of the distances between distinct points. 

  1. How are classification and regression different? 

Regression analysis is used for working with constant statistics, such as estimating markets at a given point in time. Classification is used to generate specific results and categorize details. 

  1. How do you confirm that the model is not overfitting? 

Ensure that the model’s design is basic. Reduce model noise by using less variables and parameters. Cross-validation methods such as K-folds cross-validation help in maintaining a healthy level of overfitting. Regularization methods such as LASSO help in preventing overfitting by penalizing particular criteria that are vulnerable to overfitting. 

  1. What do the terms, Training Set and Test Set mean? 

We divided the provided data set in two distinct sections: ‘Training Set’ and ‘Test Set’. 

The term ‘Training Set’ refers to the subset of the dataset utilized to train the model. 

The term ‘Testing Set’ refers to the dataset’s subset used to evaluate the learned model. 

  1. What are the primary benefits of Naive Bayes? 

In comparison to models such as logistic regression, a Naive Bayes classifier converges easily. As a consequence, in the case of a Naive Bayes classifier, we need less training data. 

  1. What is ensemble learning? 

Ensemble learning generates numerous base models such as classifiers and regressors and then combines them to produce improved performance. It is used to construct correct and self-contained item classifiers. There are two types of ensemble methods: sequential and parallel. 

  1. Define dimension reduction in ML.

It is the method of shrinking the function matrix in size. We attempt to minimize the number of columns to obtain a complete feature set, either by clubbing columns or by eliminating unnecessary variables. 

  1. What do you do if the model has a low bias but a large variance? 

When the model’s expected value is very similar to the real value, this is referred to as low bias. In this case, bagging algorithms such as the random forest regressor may be used. 

  1. Distinguish the random forest algorithm from the gradient boosting algorithm. 

If random forest makes use of bagging techniques, GBM makes use of boosting techniques. 

Random forests are mostly used to minimize variance, while GBM is used to reduce both the bias and the variance of a model.

4) Python interview questions for data science

Data science goes beyond fundamental data processing and necessitates proficiency in more sophisticated methods. Thus, if you deal with large amounts of data that need complicated computations or the development of aesthetically appealing and interactive graphs, Python is one of the most powerful solutions available. Some of the most important Python data science interview questions are listed below

  1. What is the purpose of a Python dictionary? 

In Python, a dictionary is one of the built-in data types. It specifies an unordered mapping of specific keys to their corresponding values. Dictionaries are indexed using keys and values may be any legitimate Python data form (even a user-defined class). Notably, dictionaries are modifiable, meaning they could be altered. Dictionaries are constructed using curly brackets and they are indexed using the square bracket notation. 

  1. Distinguish between lists and tuples. 

Both lists and tuples are composed of objects, which are every Python data type’s values. However, these two data forms vary in many ways: 

  • Tuples are immutable, while lists are mutable. 
  • Square brackets denote lists, while parentheses denote tuples. 
  • Lists execute more slowly than tuples do. 
  1. Explain lambda functions. 

In Python, lambda functions are anonymous functions. They’re instrumental when you need to describe a feature that’s just one expression long. Thus, rather than formally specifying the small function with a particular name, body and return expression, you may use a lambda function to write it in a single short line of code. 

  1. Explain the usage of list comprehensions in Python. 

List comprehensions allow the production of lists in a succinct manner. 

Traditionally, square brackets are built using a list. However, when a list comprehension is used, brackets include a phrase preceded by a ‘for’ clause and, if appropriate, ‘if-clauses’. When the phrase provided is evaluated in the sense of these ‘for’ and ‘if clauses’, a list is generated. 

Python programming is not going anywhere, so it is essential to emphasize data science interview questions python while preparing.

  1. What precisely are pandas? 

Pandas is a free and open-source Python library that offers fast and scalable data structures and visualization methods that render dealing with relational or labeled data intuitive and straightforward. 

  1. Which Python libraries do data scientists use to visualize data? 

Matplotlib is the primary library used in Python for plotting results. However, plots produced using this library need significant fine-tuning to appear polished and professional. Consequently, many data scientists favor Seaborn, which facilitates the development of visually pleasing and meaningful plots utilizing a single line of code. 

  1. Mention a few well-known Python libraries for data analysis. 

If you’re using Python for data analysis, you’re almost always going to use: 

  • NumPy
  • Pandas
  • Matplotlib
  • Seaborn
  • SciKit
  1. What data types does Python support? 

Python contains the following data types: 

  • Number (float, integer)
  • String
  • Tuple
  • List
  • Set
  • Dictionary

5) Deep learning interview questions for data science 

Deep learning is one of the critical areas of computer technology right now. It is a collection of techniques for predicting outputs from a layered series of inputs. Companies worldwide embrace deep learning, and everyone with tech and data expertise will find many career openings in this area. A career in data science has the opportunity to be the most exciting work you’ve ever had. However, you might want to brush up on your deep learning skills before applying for a data scientist role. Some of the top data science interview questions in the domain of deep learning are listed below.

  1. What is a neural network? 

Neural networks are a condensed version of how humans understand, influenced by the way neurons in our brains function. 

The most used neural networks have three network layers: 

  • An Input Layer
  • A Hidden Layer
  • An Output Layer
  1. What is normalisation of data and why do we need it? 

The term, data normalisation, refers to the task of standardizing and reforming data. This is a pre-processing phase that eliminates redundant data. Frequently, data is received in several formats. In these instances, you can rescale values to match into a defined range, thus improving convergence. 

  1. What precisely is a Boltzmann Machine? 

A Boltzmann Machine is a fundamental deep learning model that resembles a simpler variant of the Multi-Layer Perceptron. This model consists of a transparent input layer and a hidden layer — it is a two-layer neural network that allows stochastic decisions to switch on or off a neuron. Nodes are bound through layers but not between nodes inside the same layer. 

  1. What are the roles of activation in a neural network? 

At the most fundamental stage, an activation mechanism dictates whether or not a neuron can fire. It takes as input to any activation function the weighted sum of the inputs and bias. Activation mechanisms include the step function, Sigmoid, ReLU, Tanh and Softmax. 

  1. How is the cost function defined? 

Often known as “loss” or “error”, the cost function is a metric used to determine your model’s efficiency. A cost function is used to calculate the output layer’s error during backpropagation. We reverse the error across the neural network and use it for the various training functions. 

  1. How is gradient descent defined? 

Gradient descent is the ideal algorithm for decreasing either the cost function or an error. The goal is to locate a function’s local-global minima. This establishes the path in which the model can proceed to minimize the mistake. 

  1. What is an MLP (Multilayer Perceptron)? 

MLPs, including neural networks, consist of three layers: an input layer, a hidden layer and an output layer. Its composition is identical to that of a single-layer perceptron except with one or two hidden layers. Although a single-layer perceptron can classify only linear separable groups with binary performance (0,1), MLP is capable of categorizing non-linear classes. 

But for the input layer, each node in the subsequent layers is enabled by a non-linear operation. This implies that the input layers, the data that is sent, and the activation mechanism are all dependent on the nodes and weights being added together to generate the output. MLP makes use of a guided learning technique known as backpropagation. The neural network measures the error using backpropagation and the cost function. It propagates this mistake backward from the point at which it originated (adjusts the weights to train the model more accurately). 

  1. How are weights in a network initialised? 

There are two possible approaches here: we can either zero out the weights or distribute them arbitrarily. 

  • Through initializing all weights to zero, the model resembles a linear model. Each neuron and layer performs the same procedure, producing the same performance and rendering the deep net ineffective. 
  • Both weights are initialized randomly: In this case, the weights are initialized very near to zero. It increases the model’s precision when each neuron conducts a particular computation, and it is the most employed technique. 
  1. Explain the use of a computational graph. 

Anything in TensorFlow is constructed on the idea of building a conceptual graph. It consists of a network of nodes, each of which executes a mathematical operation. The nodes represent mathematical operations, while the edges represent tensors. Since data flows in the form of a line, it is often referred to as a DataFlow Graph. 

  1. Why is Tensorflow the most popular deep learning library? 

Tensorflow supports both C++ and Python APIs, making it easier to operate with and compiling quicker than other deep learning frameworks such as Keras and Torch. Tensorflow is compliant with all CPU-based and GPU-based computing systems.

6) Statistics interview questions for data science 

Statistical computation is the technique used by data scientists to transform raw data into forecasts and models. It is impossible to excel as a data scientist without solid experience in statistics. The most frequently asked data science interview questions in statistics are listed below.

  1. What is sampling? 

Data sampling is a statistical analysis methodology that involves selecting, manipulating and analyzing a representative subset of data points to uncover patterns and trends in the broader data collection under examination. 

  1. What is the differentiation between a type I error and a type II error? 

When the null hypothesis is valid but ignored, a type I error exists. When the null statement is incorrect but is not discarded correctly, a type II error exists. 

  1. What is statistical interaction? 

In simplest terms, an interaction occurs when the influence of one factor (input variable) on the dependent variable (output variable) varies according to its degree.

  1. What does selection bias involve? 

Selection (or sampling) bias happens in an ‘active’ context where the sample data collected and prepared for modeling has characteristics that do not indicate the model’s actual, potential population of cases. Active selection bias arises when a subset of data is omitted from the study in a systemic (i.e.  non-random) way. 

  1. What is an example of a non-Gaussian distribution in a data set? 

The Gaussian distribution is a member of the exponential family of distributions. Still, there are many more of them with similar ease of use in many situations, and they may be used where necessary if the individual doing the machine learning has a good background in statistics.

  1. How can you calculate the mean length of all fish in the sea?
  • Define the level of confidence (most common is 95 percent).
  • Take a sampling of fish from the shore (the quantity of fish should be greater than 30 for best results).
  • Determine the lengths’ mean and standard deviation.
  • Determine the t-statistics.
  • Determine the confidence interval for the mean length of all the fish.
  1. What do bell curve distribution and Gaussian distribution mean?

The normal distribution is also known as the bell curve distribution or the Gaussian distribution. It is called a bell curve since it resembles a bell. It is known as the Gaussian distribution after Carl Gauss.

  1. What are quantitative and qualitative data otherwise known as?

Numeric data is another name for quantitative data. Categorical data is another name for qualitative data.

  1. Where do long-tailed distributions come into play?

A long-tailed distribution is one in which the tail rapidly diminishes near the end of the curve. The Pareto principle and product sales distribution are two good examples of long-tailed distributions in action. It is often used in classification and regression problems.

  1. What is an outlier? How could outliers in a dataset be identified?

Outliers are data points that differ significantly from the rest of the observations in the dataset. Depending on the learning phase, an outlier could significantly reduce a model’s precision and performance. Outliers are detected using one of two methods, standard deviation/z-score and interquartile range (IQR).

7) R interview questions

R is an open-source programming language that is used for a wide range of activities and operations such as data visualization, mathematical analysis, prediction analysis, predictive modelling, data processing and so on. Here are some of the most relevant R Programming interview questions to plan for:

  1. What exactly are R packages?

Packages are sets of files, R functions and compiled code in a predefined format that are contained in a library. The user-written feature in R language is one of R’s strengths.

  1. In R, which method is used to export data?

There are several methods for exporting data into other formats such as SPSS, SAS, Stata and Excel Spreadsheet.

  1. What do you understand by GGobi?

GGobi is an open source visualization program for analyzing high-dimensional typed data.

  1. Why is a car package used?

It offers a number of regression options, such as scatter plots and variable plots as well as improved diagnostics.

  1. Which variables are denoted by capital letters?

Upper case letters are used to reflect categorical variables.

8) TensorFlow interview questions

  1. How does data get into TensorFlow?

There are two methods for loading data into TensorFlow before training machine learning algorithms:

  • Loading data into memory: The data is loaded into memory as a single array device here. It is the simplest method of loading the files.
  • Data pipeline with TensorFlow: It employs the built-in APIs to load the data and feed it to the algorithm.
  1. Where is TensorFlow more commonly employed?

TensorFlow is used in both machine learning and deep learning domains. TensorFlow, as the most important technique, has the following key use cases:

  • Time series analysis
  • Recognition of images
  • Recognition of speech
  • Video upscaling
  • Applications that depend on testing
  1. In TensorFlow, what is the distinction between tf.variable and tf.placeholder?
  • tf.variable: Defines values for time-varying variables. When defined, it must be initialized.
  • tf.placeholder: This class defines inputs that do not alter with time. It is not necessary to initialize the object when defining it.
  1. What does the embedding projector in TensorFlow mean?

In TensorFlow, an embedding projector is an object that is used to efficiently visualize high-dimensional details. Prior to visualization, it reads data from the model checkpoint register. It is used to display the input data after the model has inserted it in a high-dimensional space.

  1. What specifically is the concept of deep speech?

Deep speech is an open-source speech-to-text engine that makes use of TensorFlow. It is trained in using machine learning methods and employs a basic syntax to process speech from an input, and generate textual output on the other end.

  1. What are some of the most widely used optimizers in TensorFlow while training a model?

Many optimizers may be used depending on different variables such as learning pace, performance metric, dropout, gradient and more. Some of the more common optimizers are as follows:

  • AdaDelta
  • AdaGrad
  • Adam
  • Momentum
  • RMSprop
  • Stochastic Gradient Descent

9) Mahout interview questions and answers

  1. What is the history of Apache Mahout? When did it all begin?

The Mahout project was established by several Apache Lucene (open source search) community members who had a strong interest in machine learning and a desire for stable, well-documented, scalable implementations of basic machine-learning algorithms for clustering and categorization. The community was initiated by Ng et al.’s paper “Map-Reduce for Machine Learning on Multicore”, but it has subsequently grown to include a much wider range of machine-learning approaches.

  1. What are the characteristics of Apache Mahout?

The key characteristics of Mahout are as follows:

  • Taste is an open source project for CF initiated by Sean Owen on SourceForge and donated to Mahout in 2008.
  • Clustering implementations that support Mapreduce include k-Means, fuzzy k-Means, Canopy, Dirichlet and Mean-Shift.
  • Implementations of Distributed Naive Bayes and Complementary Naive Bayes classification.
  • Capabilities for distributed fitness functions in evolutionary programming.
  • Vector and matrix libraries.
  1. What are the various clustering options in Mahout?

Mahout supports a number of clustering algorithm implementations, all written in Map-Reduce, and each with their own collection of objectives and criteria:

  • Canopy
  • k-Means (and fuzzy k-Means)
  • Mean-Shift
  • Dirichlet
  1. What is the Apache Mahout version 1.0 roadmap?

The next big update, Mahout 1.0, will provide significant updates to Mahout’s underlying design, including:

  • Scala’s: Along with Java, Mahout users would be able to write work in the Scala programming language. Scala makes it far simpler to program math-intensive programs than Java, allowing developers to be far more productive.
  • Spark & h2o: MapReduce was used as an execution engine in Mahout 0.9 and earlier. With Mahout 1.0, users can run jobs on either Spark or h2o, resulting in major output gains.
  1. Can you explain the recommendation engine?

The recommendation engine is a subset of information-filtering systems that can predict a user’s rating or expectations for an object. We can create a quick algorithm and a scalable collaborative-filtering engine using the taste library. The following are the key components of a taste library:

  • Data model: Users, items and preferences
  • User similarity: Resemblance between two users
  • Item similarity: Resemblance between two items
  • Recommender: Offers recommendations
  • User neighbourhood: Computes and calculates a neighbourhood of users of the same category which can be used by recommenders.
  1. How many clustering algorithms does Mahout support?

Mahout supports 2 key algorithms for clustering, namely canopy clustering and K-means clustering.


Although a data scientist’s job is not straightforward, it is satisfying, and there are numerous open positions available. These data science interview questions will help you in progressing towards your dream career. Therefore, plan for the rigors of interviews and retain a working understanding of the nuts and bolts of data science.

If you are interested in making a career in the Data Science domain, our 11-month in-person Postgraduate Certificate Diploma in Data Science course can help you immensely in becoming a successful Data Science professional.