Additional Microsoft interview context that is useful to keep, but is not counted among the 118 candidate-written experiences.

Coverage: Secondhand details, duplicate follow-ups, status-only updates, pre-interview questions, and adjacent Microsoft internship roles · 2020–2026

Why these records are separate

The main count includes 118 distinct candidate-written interview experiences. The records below are still useful, but at least one important piece is missing: the author had not interviewed yet, the post repeated an application already counted, the details came through another person, or the role did not match the AI/ML scope.

Detailed secondhand interview notes

Bing Applied Scientist internship loop

This account was relayed by a friend rather than written by the candidate, so it is not counted as an independent experience. The reported final loop included:

1. Spell correction with the nearest vocabulary word. Given a vocabulary and an input word, return the closest word when the input is absent.
2. Prefix matching. Given a vocabulary and a prefix, return every matching word.
3. Reservoir sampling. Keep K uniformly selected items from a stream.
4. Distributed reservoir sampling. Extend the method across machines that do not know the global item count or complete reservoir state.
5. Tree-node level or height checks. Determine whether selected nodes share the same level, then whether they have the same height.
6. HTML element similarity. Design features for deciding whether two HTML elements are similar.
7. Majority Element. Find the value that appears more than half the length of an array in O(n) time and O(1) extra space.
8. Coin probability. Player A flips n+1 coins and player B flips n coins. Find the probability that A gets more heads.
9. Windows menu-button experiment. Decide whether moving the Windows menu button from the left to the center is a good change. Design an A/B test and estimate the required sample size.

Other discussion covered logistic regression, modeling assumptions, projects, and behavioral questions.

Duplicate posts merged into counted experiences

Data & Applied Science PhD Intern first-round timeline

One extra post added timing and team-match context but pointed back to questions already represented by the same application. It did not add a new standalone prompt.

Data & Applied Science PhD Intern background screen

Another first-round post from an application already counted described a 30-minute discussion about a prior project, teamwork, explaining deep learning to a nontechnical person, and evaluating Microsoft products. There was no standalone technical question.

Status-only and pre-interview records

1. An unspecified 2026 new-grad post mentioned a scalable URL-shortener topic, but the role was not identified as AI/ML and the interview details were incomplete.
2. A Windows-related intern post gave an interview date and later application-status updates but no interview questions.
3. A Data Scientist Intern offer post asked about return-offer rates without describing the interview.
4. A Data Scientist Intern candidate asked what to expect from a future final round before completing it.
5. A Data & Applied Sciences Intern post discussed recruiter follow-up after a final round but never added the promised interview details.
6. A Data Scientist Intern candidate asked about a disappeared application and recruiter silence after a phone screen.
7. A Hiring Event status post said other candidates received three or four work-based questions, but the author did not share their own questions and the replies could not be tied to that interview.
8. A WebXT candidate asked about a future interview and summarized the team’s work in search, recommendations, computer vision, NLP, and reinforcement learning.
9. A Data Scientist Intern candidate asked for advice choosing among Azure, Data Platform + Growth, and Windows + Devices teams before the final round.
10. One preparation post shared a third-party question document but did not describe a completed interview.
11. A BizQA Intern candidate asked whether anyone had interviewed for the role but shared no personal interview experience.

Adjacent Microsoft internship context

Explore Intern

The Microsoft Explore first round was a general undergraduate CS and behavioral phone screen. It did not contain AI/ML, data-science, applied-science, or research content, so it is kept here rather than counted in the main series.

Role not disclosed

One PhD internship report mentioned LeetCode 273 and behavioral questions, but it did not identify the team or establish an AI/ML-related role.

Final note

Parts 1–9 contain the 118 counted experiences. This appendix keeps the remaining research context visible without presenting requests, reposts, secondhand notes, or adjacent roles as additional candidate experiences. Missing details remain missing rather than being guessed.