My Amazon Applied Science Interview Journey

I applied to the position in the Last Mile Geospatial team in Bellevue City on the Amazon website around mid-October. The callback for the interview came through in mid-November, and the first phone screen round was scheduled just before the Thanksgiving break. Making it past the initial round, I received the call for the second phone screen round, which took place in the second week of December. Following this, I was then called for the last set of virtual onsite rounds. Due to the holiday season, the interview was conducted after New Year’s in the second week of January. So, the entire process consisted of 8 interview rounds spanning roughly two months.

Let’s delve into the key components:

Science Breadth:

In the ML Breadth round, the focus was on assessing the depth of my understanding across machine learning concepts. I encountered a mix of theoretical questions and practical scenarios related to applied science at Amazon. It tested my ability to grasp a broad spectrum of ML topics, showcasing the importance of a well-rounded foundation in machine learning. This would include topics in supervised and unsupervised learning — KNN, logistic regression, SVM, Naive Bayes, Decision Trees, Random Forests, Ensemble Models, Boosting, Feature Engineering, Overfitting, Regularization, Dimensionality Reduction, best practices for hyperparameter tuning, Evaluation metrics, Regression, Clustering, Neural Networks, RNNs, CNNs, Transformers.

Science Depth:

The Science Depth segment involved a resume deep dive, where detailed questions probed into my past work experiences. This round aimed to uncover the depth of my expertise in specific areas, emphasizing the practical application of my knowledge. This would entail understanding the tradeoffs made during the project, the different design decisions, results and impact on the organization and understanding how successful was the project at solving the problem at hand using business metrics if required. Nitty gritty details of implementation are enquired during the interview and its important to take a look at past projects and know every little detail of it and study its impact.

Science Application:

The Machine Learning Case Study in the domain of the job role provided a practical challenge to assess my ability to apply theoretical knowledge to real-world scenarios. This segment gauged my problem-solving skills within the context of the job, giving me an opportunity to showcase my ability to translate theoretical concepts into actionable solutions. This would entail first understanding the business problem, and then methodically come up with steps for problem formulation and a solid reason to go for a machine learning based solution. The next part would be to come up with the data collection, feature engineering and talk about the different machine learning models and finally talk about evaluation metrics, training strategies and understanding the business metric and A/B testing the model to understand feasibility for replacing the existing model.

Leadership Principles:

The Behavioral Style questions in the Leadership Principles round were designed to evaluate my alignment with Amazon’s core leadership principles. Through scenarios drawn from my past work experiences, I was assessed for various leadership skills. This round, often conducted by a bar raiser, held significant importance in determining my suitability for the role, underscoring Amazon’s commitment to strong leadership qualities. A strong emphasis is given on the STAR format — Situation, Task, Action and Result hence it’s very important to follow this structure when answering any scenario based question.

Coding:

The Coding segment comprised LeetCode-style Data Structures and Algorithms questions. This component tested my coding proficiency and problem-solving abilities. Topics would include Data Structures — Arrays, Hash maps, Graphs, Trees, Heaps, Linked List, Stack, Queue and algorithms — Binary Search, Sliding Window, Two Pointer, Backtracking, Recursion, Dynamic Programming, Greedy. Data Manipulation libraries — Pandas and SQL. Coding concepts from Machine Learning, Probability and Statistics.

Tech Talk:

An intriguing component of the interview process was the Tech Talk, a platform for me to showcase one of my previous projects. This session involved a 45-minute presentation, allowing me to delve into the details of the project, its objectives, methodologies employed, and, most importantly, the outcomes achieved. This presentation was a chance to demonstrate my communication skills, presenting complex technical information in an accessible manner. Following the presentation, the last 15 minutes were dedicated to a Q&A session facilitated by the panelists.

Conclusion:

The interview overall went smoothly, and I performed well in both the machine learning and coding rounds. However, I faced some challenges in articulating my responses to the leadership principles, and my answers didn’t quite meet the desired standard for scenario-based questions. Despite feeling disappointed about receiving a rejection after enduring 8 intensive interview rounds, I’ve chosen to view this experience as a valuable learning journey. I am committed to continuously working on enhancing my skills and refining my approach for future opportunities.