I have been at both sides of the table in the interview process in last 5 years – as an interviewee and as an interviewer. Therefore, I am eligible to give you a clear picture of what happens at both sides of the table. In this article, I would like to explain you – how can you make a good impression on the interviewer and get selected in the interviewer. I would throw light on some of the critical aspects if taken care of properly, can increase chances of your selection significantly. The interviewer will have to think twice before rejecting you even if you are not much strong technically.
Let’s be very clear in the starting of this article only that there are certain job opportunities where employer looks for very specific one or two skills. In such cases selection chances are binary. If you have that skill and showcase it well, you will be selected and if you don’t have the skill, you will be rejected. It’s as simple as that. You can’t do much about it. But good thing is that such jobs are only 5-10% in data science field. Most of the jobs in year 2019 and following few years would be generic in nature as most of the companies are still exploring the data science and how they can benefit from it. The employers are looking for data scientists who have some knowledge of most of the widely known concepts and algorithms. They are looking for generic data scientists who have exposure of important algorithms and can be trained on something specific later. They know if you have knowledge of classification then you can do good in clustering also after some learning. This article is about such 90% generic data science jobs. I would illustrate all points which made impression on me as an interviewer about the interviewee. All this is coming from my personal experience as an interviewer when I selected and rejected interviewees, and also as an interviewee when I got selected and rejected by interviewers.
This article is both for experienced candidates and freshers. Therefore, when I say Project, it means different for both. For an experienced candidate, it means the project done by him/her in the company where he/she works. For a fresher, it means the project he/she did during the course he/she joined to learn data science. The following aspects are very critical which will make you stand out among other applicants and help you crack any data science interview :
- Attire and body language
- A good enough resume
- Communicating the problem statement of your projects well
- Complete knowledge about the entire project
- Clarity about the algorithms you used
- Data cleaning and preparation
- Final results and recommendation – performance parameter
- It’s all about story
- Play in your strong zone
1. Attire and body language
I will not focus on this point much as you have already read about it a lot, but that does not take away the fact that its as important as rest of the points. Always wear formal shirt pant and formal watch when you go for interview even if your interviewer comes in half pants. Once you join the organization, you can adopt their culture and wear what other employees wear. But my sincere advice is that play safe and wear formals during interview. Also carry a smile on your face, no matter how nervous you are. Just take it easy. Be confident and rest assured that the worst thing that can happen to you in that you will get rejected in interview, but with an experience of attending an interview. You will come to know what all questions are being asked in interview these days. Where do you lag? What are areas you need to improve in yourself or your resume. These things you will never come to know sitting at home. You have to attend an interview for that. Be sincere but don’t take interview too seriously by thinking what will happen if I get rejected and all. Keep applying to many companies. Don’t get attached to just one company. At the end of the day there is only one person who is most important and whose opinion matters most and that person is you. All that matters is what you think about yourself, not what an interviewer thinks of you. I have seen some candidates sweating because of nervousness in interviews. Don’t do that. There are hundreds of other opportunities out there and you need only one. Be yourself. Be cool.
2. A good enough resume
A good resume sets up the platform for you even before interviewer sees you. A professional looking resume talks about you. It gives a mental picture of the candidate to the interviewer. In around 50% of the cases, interviewer can correctly judge a student by looking at the resume. Even if an interviewer does not like to prejudge a candidate before interviewing but it inherently happens that by going through a resume, the interviewer creates an image of the candidate and that impacts the interview result to a certain extent. That’s why it’s important to have a good neat and professional resume. You resume doesn’t need color, highlights or images. Just plane simple text giving an overview about your projects, tool and technologies is more than enough. Read here how to create a good resume.
3. Communicating the problem statement of your projects well
This might sound very trivial to you but believe me most of the candidates are not able to do that. They are either not very clear about the problem statement themselves or are not able to communicate it well to the interviewer. Many a times interviewer and candidate are on different pages while discussing about the project done by the candidate. These situations are not good for the candidate when the interviewer is struggling to make a mental map of the project all the time during the interview. All your hard work will go in vain if you are not even able to make the interviewer understand about your project on which you work for months.
All your approaches, algorithms or results come later; your first task is to make the interviewer understand about your project. To overcome this issue, try explaining your project to your family member or a friend from non data science field. If your family member or friend is not able to understand then that will be a great wake up call for you. Once you succeed in explaining your project to your family member, you can explain it to any one including interviewer.
Problem statement has to be in Business context, not in technical context.
This is noteworthy that your problem statement has to be in business context and not in technical context. Your problem statement should only talk about business. It does not make any sense to any one if your problem statement says :
You wanted to find two categories – good and bad, by using logistic regression
I have personally seen candidates saying this about the problem statement :
The task was to apply classification algorithm to predict 0 or 1
For god sake, don’t do this. Please understand the business context well behind your data science project well. Sometimes it happens that there is not much visibility of business in the company for data scientists. They just do as they are told. If you are just satisfied with doing what is told to you and not asking why, then you are doing wrong to yourself, my friend. It will not only harm your prospect in professional life but in personal life. Your why needs to be very clear. You need to ask your manager about the business context of the data science problem you are working on. This will not only improve your chances to clear a data science interview but also leave a very good impression on your manager that you are not the “frog in the well” and you want to see the big picture. Gaining some business knowledge might help you in getting promoted also in your current company.
A simple business context problem statement would look something like this :
Bank XYZ was facing a lot of frauds in its card transactions. Many a times, money involved in such fraud transactions is used in illegal activities like terrorism, drug trafficking, etc. It led to huge fines on the bank by auditors and government. Our task was to find fraud transactions as soon as they take place so that remediation steps can be taken before much damage is caused due to them.
4. complete knowledge about entire project
This is something you must have even if you are not appearing for the interview. Data Science project in itself is a journey spanning over months. There are multiple stages of every data science project, starting from defining problem statement, getting the data till finally delivering results. In reality, you may not be part of each stage of the project. But the truth is that you are expected to know about each stage. Saying simply that you don’t know anything about data cleaning stage because it was done by others and your task was to train the model does not leave a very good impression. Even interviewer knows that you can’t be part of each stage of the project but you need to know about it up to a certain extent. You should be able to give an overview of what interviewer is asking for and if he wants to know more, you can politely tell him that as you were not part of the effort, you have only this much information.
To get complete knowledge about the project, just go to the guys who have worked on different parts of project and get some idea. Generally, cross team meetings or large team meetings happen at regular intervals in companies to know each other’s work well. But even if it does not happen in your company, take a lead and get this knowledge yourself by approaching the concerned persons. It is always give and take. When you ask for some knowledge from others, then you also tell them about your work on the project.
5. Clarity about the algorithms you used
You need to know the algorithms you used in the project well. It is expected that if you have applied an algorithm in your project, you must have sound knowledge of that algorithm. Prepare answers to justify why you applied this algorithm only, why not other algorithms. Be clear not only about application of the algorithm to your dataset but also about the theory behind algorithm
In general, you should be aware about these three things about the algorithm:
- Assumptions of the algorithm if there are any – when it can / can’t be used
- Advantages of the algorithm
- Disadvantages of the algorithm
If you know these answers, you leave a very good impression on interviewers that you know what you are doing and what you are doing. It automatically brings you ahead of 99% of candidates. Don’t ever give answers that you used a particular algorithm because your senior asked you to.
6. Data cleaning and preparation
This is somehow the favorite topic of many interviewers because it’s a known fact that the data that you get for your project can not be used for training the model as is. Data is inflicted with many data diseases like outliers, anomalies and sometimes data needs to be prepared or transformed for better prediction results. I have personally faced these questions in almost all interviews I appeared for. Prepare a good answer of these questions that what all outliers, anomalies, bad values you found in the dataset and how you cleaned the data.
Some of the examples of such issues and approaches to handle them could be as following:
For what issue, what approach will you adopt is your call. If you think, you have a lot of data and you afford to lose such rows then remove these rows. If you already have less data then you need to impute such values by some acceptable values. I will write a complete article about data cleaning and data preparation strategies in future. Don’t undermine the importance of this step of data science project as it’s one of the favorite steps of interviewers.
7. Final results and recommendation
This is crucial phase of data science project where you can influence the interviewers. Here you actually complete the circle and connect the dots. – What was the problem statement, why and how you started the data science project, what were your results and how these results can successfully solve the problem.
Start with explaining the results of your model to your interviewer. Sometimes you need to convert these results into simple laymen terms or charts to describe them to your clients. Explain the interviewer why and how you transformed model results into interpretable conclusions for comprehensible illustration. Sometimes, results are converted into recommendations for clients. Describe how your recommendations helped the client significantly. If you have quantitative numbers to show that then it’s best, like, implementation of your recommendation reduced the problem by 10%. If you don’t have this number, you can tell the interviewer that reduction was significant but you don’t have exact numbers or the client is in the process of implementing the recommendations and we are waiting to hear from client.
Let’s take a very simple example here for better understanding. Let’s say your client came to you with a problem that the employee’s loyalty is very less in his company. Employees tend to leave company withing couple of years after joining. Your client invests a lot of money in hiring and training employees but when employees leave just within 2 years, it’s a loss for company. Company wants you to give then recommendations that what kind of employees should they hire so that they stay longer with company. So, you do modeling with the historical data of company employees and see that employee married is coming as very important variable and its has positive relationship with your target variable i.e. length of stay with company. So, you will interpret the result in laymen term and add this result into recommendations to client that married employees are more likely to stay with the company for longer duration than unmarried. Therefore, if you have two candidates with equal caliber, one married and other one unmarried, and you have only one position then go with married candidate. After implementing this suggestion, employee’s attrition rate within first two of joining reduced by 5%.
8. Play in your strong zone
Lastly, just keep this in mind that interviewers are there to know what all concepts you are good at or you have in your resume, and not to know what all concepts you don’t know. In view of this, you have to show the side you are strong at. Sometimes, interview goes in some directions which are not your forte. It’s your duty and skill to bring the interview back to your strong zone. For example, you have worked on few good basic classification machine learning algorithms and have got good knowledge, but interview suddenly goes into deep learning area. Let’s you have very vague idea about Deep Learning but not good enough to face any interview question then your task is to bring the interview back to basic classification machine learning algorithms. There are two ways to do that :
- Re-introduce your forte in the discussion
- Say sorry to interviewer politely
1. Re-introduce your forte in the discussion
Just an example – You can tell the interviewer that it’s correct but you think that your machine learning algorithm can also give the almost equally result without overhead of costly computing resources and time consumption in running the algorithm.
2. Say sorry to interviewer politely
Just tell the interviewer politely and confidently that you have heard about the concept he is asking for. But you have no working knowledge of it so won’t be able to answer questions. Honest is best policy even today.
9. It’s all about story
Data science is a science but it’s an art equally. Stories are very powerful, and they are able to make the connect with the audience quickly and leave long lasting impacts. People like stories and interviewers are no exception. They also love people who are good at telling stories. Nobody likes a dud technical interview. Even interviewers are humans and they also need to interview candidates in good, relaxing and entertaining environment. You need to build a good story around your project. In fact, your project has to be told in form of a story. This story might include:
- How it all started – what was the pain point of your client?
- What all you tried to get away with the pain point
- Why you decided to go ahead with the
- What all challenges you facedtodealwith
- higher management/client
- data gate-keepers to get the data
- other stakeholders
- How you carried out the entire project weathering all challenges
- How you presented the results to get buy-in of your management/client and how happy & surprised they were to get the hidden insights from the data which they were unaware of even after being with company for years.
- How things turned around after applying your recommendations
And, answers to these questions don’t need to be technical. Simple and plain English will work wonders. All of the above points might not be relevant in your case, but you must have got the idea what I am trying to convey. It’s your story – build it yourself.
Having explained all the key aspects of getting a candidate’s selection chances much stronger, there are many other factors which are beyond candidate’s control also play crucial role in getting the final job offer. Even after doing good in job interview by following all the above points, you may fail to get the offer because of no mistake of yours. It might happen that while taking your interview, position was open but later company dropped the position because of some internal strategical changes. It might also happen that the company’s finance department does not approve the salary offered to you by HR department. These all are the things which I have also experienced as a job candidate. Don’t get disheartened because of getting rejected for these issues, or rather any issues. Just keep and learning & improving your skills, and always keep this in my mind there is no shortage of good opportunities in good companies. Just chill. Go for a movie, enjoy and then get back to sharpening your data science skills. More relaxed and cheerful you are, more are the chances of you getting selected 🙂