6 steps to crack Data Science Interview and get your first Job as (with or without Job experience)
I have personally used these techniques to crack interviews
I still remember my struggle to get my first job as a Data Scientist. Let me tell you it’s not at all easy to get into the field.
Most positions need you to have some experience for them to consider you for an interview. So it’s a vicious circle.
But by following some of the tips mentioned below you can surely get your first job as a Data Scientist.
In case you are interested in knowing my background, here is the link to my transition story:-https://medium.com/illumination/key-takeaways-for-successful-transition-to-data-science-from-my-journey-da0cb0215f31
A successful salesman always puts the customer’s perspective at the core of his sales strategy.
Similarly, as candidates, we need to think from the interviewer’s perspective and resonate with what they are looking for.
All the below-mentioned points will help you think in this direction
- Do Projects that are related to your field:-
I have taken hundreds of interviews of freshers/transitioners, and one thing that puts me off is when I see the same old projects in candidates’ CVs.
Most of the candidates either use Titanic data, Boston housing data, Iris data, etc…
Either pick up some projects that are relevant to your past experience or if you are a fresher do some research on which fields have the most openings and try to work on projects from those fields.
Some of the fields having lots of openings that I have come across are Banking, Insurance, Retail, Telecom
2. Speak about projects that are related to the field you are interviewing for:-
The most common question in a Data Science interview is “tell me about a project that you have worked on”.
This is your time to shine.
You see Data Science/ Machine Learning is a vast field, and no one knows what will one be asked. But this is one question where the ball is in your court.
If you have done your research on what the company does here you can showcase your knowledge and prove how you will be a good addition to the team.
Example:- If the client works in the Insurance domain, speak about the project related to Insurance.
Ex:- Propensity to surrender
Explain why you picked up this as a project and how this can help the company know well in advance who are the customers who might surrender and use our predictions how the sales folks can reach out to those customers and retain them.
Put in some figures Ex :- Considering 20% of the customers have the propensity to surrender and even if we are able to prevent 50% of those, that will translate into these many $ savings etc…
This shows the interviewer that you have communications skills and you can speak to business in a language they would understand.
3. Research about the interviewer:-
This is very very important.
Knowing your audience is half battle won.
If the interviewer is a technical guy, then approach your interview that way.
Explain your project from a technical angle. Explain the data set how you worked on getting the data, pre-processing, data engineering (how this is the most important), one hot encoding, modelling, evaluation techniques etc etc…
Go in full depth to explain the models you used. Keep things very technical. Learn the most commonly used models explain them in detail- this is import as everyone is aware of common models and can judge your answers and then throw in new models which can also be explored.
But if the interviewer is a Functional Manager (mostly the second round after technical round)- Then discuss things from a 50 feet view (unless they specifically ask technical questions)
You should be able to explain things in a lay man’s language. This is expected when Data Scientist speak with Non-Technical folks.
4. Explain projects end to end:-
Many candidates start with we got data from Kaggle. In the real world, this rarely happens.
Firstly to have to get the data from DBs or somewhere else, secondly the data is never clean.
So start with projects that require you to scrap data from the web or get data by joining data from many sources etc…
Try to speak with friends working in the industry and understand how they do things.
Try to keep things as close to real life as possible
5. Honesty is the Best policy and Show willingness to learn:-
No one and I mean No one has answers to all the questions. When you are confronted with any such situation, accept that you do not know rather than giving wrong answers and say things like this is something new to me. I would love to learn and implement something like this.
6. Research about model explainability:-
As per statistics 85% of the models developed never move into production and the others that go into production seldom get used.
One of the main reason is business doesn’t have the confidence to go live.
Explore global feature explanation techniques like Feature Importance and more importantly learn and explain local feature explanation techniques like SHAP values.
Believe me this is a game changer. Many senior Data Scientists don’t know much about things like SHAP.
Feature importance explain which are the most important features on a dataset level.
SHAPley values explains which feature contributed the most in predicting.
So Image you are building a Loan Repayment Model.
If you can explain based on which features our model is rejecting to give loan to a specific customer. The business will surely consider our model.
Hope you enjoyed this article. If you did do consider following me as I intend to write such practical articles that will help the community.
