2020–2026 Microsoft CoreAI and AI/ML Intern OA / Interviews: 118 Experiences (Part 2 :)

This part preserves 15 Microsoft AI/ML internship reports from 2024–2026.

Coverage: 2024–2026 · candidate reports 005–019 · phone screens and final loops

Series note: Part 1 focuses on the CoreAI / AI-ML Software Engineering Intern loop. This part preserves other Microsoft AI/ML internship reports without treating them as CoreAI SWE evidence.

2026 reports

005: Copilot Research Intern — phone or technical screen

Report date: 2026-09-27 · Role: Copilot Research Intern · Stage: Phone / technical screen

Microsoft Copilot Research Intern report describing a pure coding video interview, one exact LeetCode problem, interviewer hints, two solution families, complexity discussion, and a pass result.

1. Top K Frequent Elements. The exact task is LeetCode 347, Top K Frequent Elements. A hashmap-plus-heap solution is O(n log k), while bucket sort can be O(n). The interviewer may ask complexity and optimization follow-ups.

006: Applied Science PhD Intern — phone or technical screen

Report date: 2026-02-23 · Role: Applied Science PhD Intern · Stage: Phone / technical screen · Interview date: 2026-02-19

Detailed Microsoft Applied Science PhD Internship Opportunities - Redmond report with one initial screen, three back-to-back interviews, project and ablation-study discussion, NLP/ML fundamentals, two coding tasks, RAG design, and an offer.

1. Return Every Pair with the Minimum Numeric Difference. Given [-1,-2,-4,-5,7,10], find the minimum difference and return every pair achieving it. The reported output is [[-5,-4],[-2,-1]]. The interviewer described the task as Easy.
2. Compute Dog Retrieval Precision and Recall. There are ten image files, each containing an animal image. An API returns k files; compute precision and recall for dog images. Handle abnormal API results such as None or a result count different from k.
3. Design a Retrieval-Augmented Generation System. Design a RAG system. The candidate discussed an agentic-tool framing.

Other reported topics or limits: Ablation Design, RL, Agents, Metrics, and Regularization.

007: Applied Scientist Intern — final interview loop

Report date: 2026-02-13 · Role: Applied Scientist Intern · Stage: Final loop / onsite

Microsoft Applied Scientist internship report covering a phone screen and three virtual-onsite rounds with ML/LLM fundamentals, from-scratch K-means, hypothesis testing, Python debugging, project discussion, case study, and an offer.

1. Implement K-Means from Scratch. Implement K-means from scratch. Discuss the algorithm's time complexity.

Other reported topics or limits: ML and LLM Fundamentals Phone Screen; Business Hypothesis Testing and a Simple Python Debug; Project Discussion and Open-Ended Case Study.

008: Data Scientist PhD Intern — phone or technical screen

Report date: 2026-02-04 · Role: Data Scientist PhD Intern · Stage: Phone / technical screen

Microsoft PhD Data Scientist internship report with an OA and three back-to-back interviews.

1. Show an Organization Chart One Layer at a Time. Model the organization chart as a tree and show each complete layer from top to bottom. Breadth-first search with a queue is the expected traversal shape.
2. Design Counterfactual Evaluation When A/B Testing Is Unavailable. Explain how to evaluate a change when an A/B test cannot be run. Discuss counterfactual-estimation approaches.
3. Diagnose a Sudden Product-Metric Drop. Choose metrics for evaluating whether a product is good. If a core metric suddenly drops on one day, determine possible causes and how to test them.

Other reported topics or limits: Two Unnamed One-Hour OA Questions; Unnamed SQL Question; ML Experimentation and Evaluation Topics.

009: Data Scientist PhD Intern — phone or technical screen

Report date: 2026-01-05 · Role: Data Scientist PhD Intern · Stage: Phone / technical screen · Interview date: 2025-11-03

Microsoft Data Scientist PhD internship report with a HackerRank OA, three back-to-back interviews, LLM-heavy research context, and an offer.

1. Implement a Precision-Recall Curve. Code or draw a precision-recall curve. The interviewer gave substantial hints after the candidate checked the definition of recall.

Other reported topics or limits: One Unnamed LeetCode and One Unnamed Data-Processing OA Task; KNN, PCA, and LLM Fundamentals.

2025 reports

010: Applied Scientist PhD Intern — phone or technical screen

Report date: 2025-12-20 · Role: Applied Scientist PhD Intern · Stage: Phone / technical screen · Interview date: 2025-11-20

Microsoft Seattle Applied Scientist Intern 2026 Summer report covering the application timeline, ML and LLM fundamentals, an unspecified easy Python string task, a rejection, and an author clarification that the candidate was a PhD student. It is admitted to the expanded AI/ML Experience corpus; problem ingestion remains outside this article-only scope.

1. No standalone technical prompt was preserved. This entry is retained for its process, timing, role, or interview-format evidence.

Other reported topics or limits: ML and LLM Foundations for Real-World Deployment; Unspecified Easy Python String Task.

011: Applied Scientist PhD Intern — phone or technical screen

Report date: 2025-11-10 · Role: Applied Scientist PhD Intern · Stage: Phone / technical screen

Microsoft Seattle Applied Scientist Intern 2026 Summer report covering an unusually direct technical first round, a deep research-project and ML evaluation discussion, two behavioral questions, advancement to a final round, and eventual rejection. It is admitted to the expanded AI/ML Experience corpus; no standalone technical problem is disclosed.

1. No standalone technical prompt was preserved. This entry is retained for its process, timing, role, or interview-format evidence.

Other reported topics or limits: Deep Research-Project and ML Evaluation Discussion.

012: Applied Scientist PhD Intern — phone or technical screen

Report date: 2025-10-28 · Role: Applied Scientist PhD Intern · Stage: Phone / technical screen · Interview date: 2025-10-17

Microsoft Applied Scientist PhD internship report covering a 30-minute initial screen, three 45-minute final interviews, DS/LLM preparation mismatch, and an offer. The complete author-written evidence is preserved for the expanded AI/ML Experience article; separate problem ingestion remains outside this article-only scope.

1. No standalone technical prompt was preserved. This entry is retained for its process, timing, role, or interview-format evidence.

Other reported topics or limits: Explain How ChatGPT Is Trained; Three DS-Focused Problem-Solving and Coding Interviews.

013: Data Scientist / Research PhD Intern — phone or technical screen

Report date: 2025-06-18 · Role: Data Scientist / Research PhD Intern · Stage: Phone / technical screen · Interview date: 2022-10-10

The 2025 post is a retrospective of a September–October 2022 Microsoft PhD Data Scientist/Research internship loop.

1. Retain the Top 1,000 Values from a Large Integer Array. Given a large integer array, find the top 1,000 values.
2. Count Set Bits with a Precomputed Byte Lookup Table. Count the number of one bits in a byte from 0 through 255. Extend the approach to an array of one million numbers by precomputing results for all byte values.
3. Design CTR Prediction from Query, Ad, and Click History. Given many historical queries, many ads, and click observations for each ad, frame an ML model that predicts click-through rate.
4. Improve a Black-Box Predictive Model. Given training data and a black-box predictor f(X)=y, explain how to improve the model. Consider dimension reduction or feature selection, kernel methods or interactions, ensembles, personalized modeling, and data augmentation. Discuss which approaches can overfit.

Other reported topics or limits: Regression, CNN Parameters, and Attention; A/B-Test Sample Size and Exploration–Exploitation.

014: Applied Scientist Intern — phone or technical screen

Report date: 2025-05-07 · Role: Applied Scientist Intern · Stage: Phone / technical screen · Interview date: 2024-11-01

Microsoft Applied Scientist internship report for the 2025 recruiting cycle, covering three coding interviews, resume questions, and behavioral discussion. It is admitted to the expanded AI/ML Experience corpus, while individual problem ingestion remains outside this article-only scope and the task identities are underspecified.

1. No standalone technical prompt was preserved. This entry is retained for its process, timing, role, or interview-format evidence.

Other reported topics or limits: Two Unnamed Array Problems Using Two Pointers and a Sliding Window; Write an Unspecified String-Processing Function.

015: Applied Science Intern (Health & Life Sciences) — final interview loop

Report date: 2025-02-28 · Role: Applied Science Intern (Health & Life Sciences) · Stage: Final loop / onsite

Microsoft Health & Life Sciences Applied Science PhD internship final-round report from Montreal. Three research-talk interviews focused on the candidate's PhD work, several LLM/ML fundamentals, deep project challenges, and problem-solving experience.

1. No standalone technical prompt was preserved. This entry is retained for its process, timing, role, or interview-format evidence.

Other reported topics or limits: Research Talk, LLM Fundamentals, and Project-Depth Discussion.

016: Student Researcher, LLM for Code — phone or technical screen

Report date: 2025-02-10 · Role: Student Researcher, LLM for Code · Stage: Phone / technical screen · Interview date: 2024-12-01

Microsoft LLM-for-code Student Researcher report describing a referral, two interviewers, resume and research discussion, one unnamed easy LeetCode task, a pass/offer path, and the author's clarification about part-time hourly pay. It is admitted to the expanded AI/ML Experience corpus with the program distinction preserved.

1. No standalone technical prompt was preserved. This entry is retained for its process, timing, role, or interview-format evidence.

Other reported topics or limits: Unspecified Easy LeetCode Task in a Research-Heavy Screen.

2024 reports

017: Data Scientist Intern — phone or technical screen

Report date: 2024-11-15 · Role: Data Scientist Intern · Stage: Phone / technical screen

Microsoft Data Scientist internship report from a CS PhD candidate, covering three back-to-back rounds, behavioral questions, one detailed median question with an optimization follow-up, and one Medium LeetCode task per round.

1. Find a Median with a Memory-Constrained Follow-Up. Find a median using a min-heap and max-heap. When memory is constrained, use Quickselect as the follow-up approach.

Other reported topics or limits: Two Other Unnamed Medium LeetCode Questions.

018: Data Science PhD Intern — phone or technical screen

Report date: 2024-10-20 · Role: Data Science PhD Intern · Stage: Phone / technical screen · Interview date: 2024-10-01

Microsoft Data Science PhD internship report for the 2025 recruiting cycle, with three consecutive interviews covering coding, ML/probability, research, and LLM design, followed by an offer.

1. Find the Median from Two Arrays. Given two arrays, find their median. The interviewer repeatedly asked for optimizations.

Other reported topics or limits: Expected Exit Time with Returning Paths; Bias–Variance, Linear Regression, Overfitting, and LLM Design.

019: AI Research Intern (non-MSR) — phone or technical screen

Report date: 2024-05-23 · Role: AI Research Intern (non-MSR) · Stage: Phone / technical screen · Interview date: 2024-05-17

Microsoft AI Research Intern pass report from a math PhD who moved into LLM reasoning and theorem proving.

1. Unnamed LeetCode Easy Live-Coding Task. The second-round interview included one LeetCode Easy problem written and reviewed live. The report did not preserve a reconstructable task contract.

Other reported topics or limits: Design How a Company Should Use AI; LLM Breadth and Research Discussion Set.

Part 3 continues with the next candidate reports.