Acing the Data Science Interview — Part 2
I wrote about preparing for the Data Science Interview — Part 1 last week.
This article is a follow up and a conclusion of this topic. Previously, I have been writing articles about AI Interview Questions for some top companies in this domain. Based on the unanimous feedback on these articles from the readers, it seemed there was a natural follow-up as to how to prepare for these interviews. In order to comprehensively cover this large topic, I divided it into parts. This is Part 2.
The next few steps are deeper into the world of AI and will require more effort and in depth learning. It should be noted that the next few steps should be prepared on an as needed basis based on the companies specific verticals and Data Science stack.
Steps 1–4: Steps to Ace the AI Interview — Part 1
5. Probability — Regression Models(Statistics — 2 days):
Probability is the likelihood that an event will occur. Likelihood that someone will buy the product online, click on the link or use a feature. All of the problems that AI can solve, it will always be at some base probability. It forms the very basis of logic that is required to build the algorithms which consume the data.
Big part of this field is preparing different models to fit the data or infer from it. Linear regression, logistic regression, gradient descent are some of the models. These models are taught in most computer science/math based basic courses. These are very important as selecting a model will help you predict where your data will go or what will happen next.
Important reading I would use:
- Practical recommendations for gradient-based training of deep architectures:https://arxiv.org/abs/1206.5533
- Online Statistics Book(Probability Section): Probability Deep Dive
- A Quick read on Regression Models: Multiple Regression
6. Machine Learning Basics (Computing — 4 days):
Machine learning is used from predicting user behaviour to self driving cars and everything in between. The intersection of cheap computational power coupled with cheap data storage has resulted into the boom of this field. It is a vast field in itself and it requires deep dives per the person’s inclination and abilities. Basics should include Tensorflow (by Google), Scikit learn which are APIs which are based on Python specifically designed for AI/ML related problems.
Important reading for preparation:
- Machine Learning Algorithms(List of the Important Algorithms): ML Algorithms
- Backdrop(Backdrop Technique in ML): Understanding Backdrop by Andrej Karpathy
- Learning Tutorials for TensorFlow: Tensorflow Tutorials
7. Advanced ML(Computing — 3 days)
Advanced ML includes various algorithms and techniques to deal with complex data problems. Time Series data can be complex to fit on a traditional ML model. Hence, LSTM — Long Short Term Memory networks are used to address this issue.
Important reading for preparation:
- LSTM(Easiest way to understand LSTMs): Understanding LSTMs
- Real-life Kaggle example for Time Series: Machine Learning for Time Series Analysis
- Advanced ML (University of Brussels Presentation): Machine Learning Strategies for Time series Prediction
8. Real World/Unstructured Data Processing(Computing — 3 days)
Google’s deep learning feat with speech recognition is well known. They have used Recurrent Neural Networks (RNNs) to achieve that goal. The network does not have enough information to proceed as the data is unstructured and hence, we include information about the sequence of characters or data and feed it to the network. This kind of RNNs are suitable for speech recognition. This is also a massive area of research and breadth and depth is dependent on the individual and the company he is interviewing at.
Important reading for preparation:
- Practical guide for using RNNs: Time-Series Prediction using RNNs
- Understanding the Back Propagation Technique: Stanford’s Back propagation Class
DataSets for AI/ML/DS Research: Comprehensive List
Once you have gotten this far, the link above can be used to consume data and get familiar with the size and the kind of work, data scientists do. Hands on practical knowledge never really goes waste.
9. Deep Learning — Neural Network(Computing — 3 days)
Deep Learning overlaps with ML in many ways. It includes concepts like GANs (Generative Adversarial Networks) and Semi — Supervised learning. CNNs or convolutional neural networks are used for image recognition and classification. This area has widespread impact in the current tech ecosystem. Depending on the company its use may vary from video generation to image classification. Hence, depending on the company this area should be prepared keeping in mind the kind of questions that might come up.
Important reading I would use:
- Convolutional Neural Networks: CS-231(Stanford Class Notes)
- Deep Learning(Book): http://neuralnetworksanddeeplearning.com/
- Advanced Deep Learning(Book): https://www.deeplearningbook.org/
This concludes the preparation for an AI/ML/DS Interview. The preparation will vary based on where the individual is approaching the field from. Preparation time for researchers will differ from data engineers to normal computer science engineers. Preparation should roughly take 20–22 days depending on the person’s existing knowledge, time constraints and company specific requirements.
Preparing for an AI Interview — Part 1: Steps to Ace the AI Interview — Part 1
Subscribe to our Acing AI newsletter, I promise not to spam and its FREE!
Thanks for reading! 😊 If you enjoyed it, test how many times can you hit 👏 in 5 seconds. It’s great cardio for your fingers AND will help other people see the story.
