Amazon

I applied for the ML Scientist / Applied Scientist Intern opportunity through the Amazon ML Challenge. The challenge was the entry point into the process, and a strong performance in the competition opened the opportunity to be considered for the internship. The initial stage was the Amazon ML Challenge, where participants had to work on a real-world machine learning problem. The challenge required understanding the dataset, performing exploratory analysis, developing a suitable ML pipeline and improving the model based on the evaluation metric. I worked through the problem by first understanding the data and identifying the important features. I experimented with different preprocessing and modeling approaches and focused on improving the validation performance rather than simply training a single baseline model. After the challenge, candidates who performed well were considered for the internship hiring process. The next stage was an online coding assessment. Online Assessment The OA consisted of two DSA problems. The questions were around the medium level and required efficient algorithmic solutions. I approached the problems by first analyzing the constraints and identifying the appropriate data structures. Time complexity was particularly important because an otherwise correct brute-force solution would not scale for the larger test cases. After clearing the assessment, I was shortlisted for the interview. DSA Technical Interview The first interview was primarily focused on Data Structures and Algorithms. The interview started with a brief introduction and then moved directly into coding. The interviewer gave me problems involving data structures and asked me to explain the approach before implementing the solution. The discussion covered concepts such as trees, graphs and dynamic programming. For the graph problem, I had to reason about traversal and determine whether BFS or DFS was more appropriate based on the problem constraints. The interviewer also asked follow-up questions around optimization and edge cases. I was expected to explain the complexity of the solution and justify why the chosen approach was correct. The interview was interactive, and the interviewer evaluated the reasoning process rather than only looking at the final code. ML + Project Interview The final round focused on Machine Learning breadth and depth along with my projects. The interviewer started by asking about my ML background and then went deep into the projects mentioned on my resume. I had to explain the complete ML pipeline, including data preprocessing, feature engineering, model selection, training, validation and evaluation. The discussion then moved into core machine-learning concepts. Questions covered topics such as bias-variance tradeoff, overfitting, regularization, model evaluation and optimization. The interviewer also asked why I had selected a particular algorithm instead of alternative models. I had to explain the trade-offs between different approaches based on the characteristics of the dataset and the objective of the problem. The discussion then moved towards deep learning. I was asked about neural networks and concepts such as CNNs, RNNs and LSTMs. There were also questions around how different architectures are suitable for different types of data. Some questions were scenario-based. Instead of simply defining an algorithm, I was asked when I would use it and what I would do if the model was performing poorly. The interviewer also went deeper into my ML project implementation. I had to explain the decisions I made during preprocessing, model training and evaluation, along with the limitations of my approach. Unfortunately, I was not selected after the final ML round. Looking back, the biggest difference between this process and a regular SWE internship interview was the combination of DSA and ML depth. The coding round required strong algorithmic problem solving, but clearing it was not enough. The final round required a strong understanding of machine-learning fundamentals and the ability to discuss projects at an implementation level. For preparation, I would divide the preparation into three areas. First, I would practice DSA, especially graphs, trees, dynamic programming, arrays and hashing. Second, I would revise core machine learning, including regression, classification, decision trees, ensemble methods, clustering, feature engineering, regularization, evaluation metrics and optimization. Finally, I would prepare ML projects in depth. It is important to understand every step of the pipeline rather than simply knowing which model was used. Questions can move from the high-level objective to very specific implementation and modeling decisions. The Amazon ML Challenge was also an important part of the experience because it provided a practical ML problem rather than a purely theoretical assessment. Overall, I wasn't able to convert the Amazon ML Challenge opportunity into an ML Scientist internship, but the process gave me a strong understanding of the expectations for an Applied Scientist/ML internship: solid DSA, strong ML fundamentals, practical project experience and the ability to reason about model choices and trade-offs.