This researcher made the right call by not canceling, and their story is worth examining closely. After multiple rejections, they finally secured an interview for a PhD applied science intern role at Amazon in the EU, and then immediately faced a crisis of confidence over the coding portion of the interview. The instinct to withdraw before being rejected again is understandable, but it is exactly the wrong move. The question was never whether to show up; it was how to prepare in a week.
The tension here reveals something important about how we think about technical interviews and the tools we use. This researcher is confident in their machine learning breadth but openly admits they have never studied data structures and algorithms, never leetcoded, and rely on AI for coding tasks. That honesty is refreshing. It also mirrors a broader shift that is worth paying attention to. When Pinterest Fine-Tunes a 4B Language Model to Remap User Journeys, it is using a model to infer user intent from behavior patterns, not to replace every engineer on the team, but to handle a specific, complex task that used to require a multi-stage pipeline. Similarly, when researchers at Amazon show how to Run PyTorch models on AWS Trainium with zero code rewrites, they are demonstrating that the barrier between domain expertise and implementation is lowering. The role of the researcher is shifting from writing every line of code to architecting solutions and knowing where to apply the right tool. That is exactly the skill set this candidate already has.
The practical takeaway is this: for anyone in a similar position, the week before the interview should not be spent cramming LeetCode problems you have never seen. Instead, focus on what you can control. Practice explaining your machine learning projects aloud, with attention to the design decisions and trade-offs. Prepare to walk through a simple data structure, like a hash map or a binary search, and show that you understand the logic even if you are not fast at implementing it from scratch. Many interviewers at this level are evaluating problem-solving approach and communication, not just raw coding speed. If you can articulate why a certain approach works and where it might break, that counts for more than memorizing the optimal solution.
This story also connects to a broader point about how we define competence in a field that is changing rapidly. The researcher says "AI does a better job than me in coding." That is not a weakness; it is an honest assessment of the current landscape. The real question is whether you can direct that AI effectively, verify its output, and integrate it into a larger system. Those are the skills that matter now. If you are reading this and hesitating to apply for a role because you do not fit the traditional coding-interview mold, consider what German researchers trace global power grid disturbances to a single cause had to do: they used deep domain knowledge to identify a pattern that automated systems had missed. That kind of insight comes from understanding the problem, not from typing faster.
The specific detail to watch here is whether Amazon's interview process for applied science roles continues to weigh DSA coding as heavily as it does for standard software engineering positions, or whether it begins to adapt to the reality that researchers bring different strengths. For now, this candidate should walk into that interview with confidence in what they know, honest about what they do not, and ready to show how they think. That is the only preparation that matters.