2020–2026 Microsoft CoreAI and AI/ML Intern OA / Interviews: 118 Experiences (Part 6 :)

This part preserves 12 Microsoft AI/ML internship reports from 2021.

Coverage: 2021 · candidate reports 063–074 · 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.

2021 reports (continued)

063: Data & Applied Science Summer Intern — final interview loop

Report date: 2021-10-27 · Role: Data & Applied Science Summer Intern · Stage: Final loop / onsite · Interview date: 2021-10-14

Detailed Microsoft Data & Applied Science summer-intern offer report.

1. Design Phone-Call Quality Classification from Limited User Feedback. Determine whether a phone call's quality is good or poor. Only a small subset of calls has user-feedback labels. Describe the solution and write pseudocode covering train/test splitting and encoding.
2. Unnamed Medium Sliding-Window Coding Task. The final onsite coding problem was an unfamiliar LeetCode Medium sliding-window problem. The report did not preserve a reconstructable task contract.

Other reported topics or limits: ML, Statistics, and Ensemble Fundamentals.

064: Data Scientist Intern — final interview loop

Report date: 2021-10-26 · Role: Data Scientist Intern · Stage: Final loop / onsite · Interview date: 2021-10-21

Microsoft Data Scientist Intern final-round pass report for a search team.

1. Implement a Precision-Recall Curve. The interviewer asked the candidate to code a precision/recall curve and discuss an optimal solution. The original author clarifies that the task required a simple manual implementation without library calls. The implementation did not need to render a graph; it only needed to demonstrate the approach and optimal algorithm.

Other reported topics or limits: Rank Passages and Choose a Bing Fact; Search and ML Concept Question Set.

065: Data & Applied Scientist Intern — final interview loop

Report date: 2021-10-23 · Role: Data & Applied Scientist Intern · Stage: Final loop / onsite

Detailed Microsoft Data & Applied Scientist master's internship onsite report for Bing Ads.

1. Count Connected Components in a Grid. The task was described as similar to LeetCode 542. Return the number of connected components in a grid.
2. Design a Text Advertising Recommendation System. Design a recommendation system for text advertisements. Cover feature construction, model choice, large-scale candidate retrieval, ranking, current ranking models, and ranking features.
3. Design a Video Recommendation System from User, Video, and Action Tables. Design a video recommendation system from user, video, and action tables. Discuss one-hot and target encoding, evolving user interests, sliding-window features and window tuning, regression versus classification, and negative sampling.
4. Predict Product Ratings from User Reviews. Given tables of user reviews, build a model to predict product ratings from one to five. Choose regression or classification; specify RNN inputs, outputs, and loss; compare RNN, GRU, and LSTM; and propose a non-deep-learning TF-IDF approach with high-dimensional sparse-feature regularization.

Other reported topics or limits: Tree, GBDT, Time-Series, Causal-Inference, and NLP Theory Follow-Ups; Two Dice Probability Decision Questions.

066: Data & Applied Scientist / Applied Scientist Intern — final interview loop

Report date: 2021-10-22 · Role: Data & Applied Scientist / Applied Scientist Intern · Stage: Final loop / onsite · Interview date: 2021-10-14

Detailed Microsoft Data & Applied Scientist internship loop for Bing Search, later transferred to Applied Scientist Intern.

1. Count Isolated Islands. The phone screen asked a simple isolated-islands coding problem.
2. Determine Whether an Adjacency-List Graph Is Bipartite. Given an adjacency list, determine whether the graph is bipartite. The expected approach used breadth-first search.
3. Score How Well a List Is Shuffled. Given a list, produce a measure of how well it is shuffled. The candidate proposed the number of monotonic sequences divided by total length. A follow-up asked for a better method, and the candidate suggested considering distances between values in monotonic sequences.
4. 3Sum. The final coding task was 3Sum.
5. Design Webpage Language Detection and Language-Aware Search Ranking. Determine which language a webpage uses. Use the detected-language information when ranking search results.

Other reported topics or limits: ML/DL, Knowledge Graph, and Research Discussion.

067: Data & Applied Science PhD Intern — final interview loop

Report date: 2021-10-21 · Role: Data & Applied Science PhD Intern · Stage: Final loop / onsite · Interview date: 2021-10-21

Detailed Microsoft Data & Applied Science PhD internship offer report.

1. Clean and Tidy Patient Vitals Data. A sample dataset contains patient ID, blood pressure, gender, heart rate, and similar fields. Explain and implement how to clean the data. Handle missing values. Split a blood-pressure value such as 120/80 into two variables. Normalize equivalent gender values such as F and Female. Discuss tidy-data structure and multiple missing-value strategies.
2. Count New-User Devices That Opened an App. Table 1 contains device_id, opened_app, and open_count; device_id can repeat. Table 2 contains device_id, isNew, and additional columns. Find how many device IDs that opened an app belong to new users.
3. Forecast Car-Model Sales and Evaluate a 2% Price Increase. A car dealer wants to predict which models will sell well and which will not. Discuss data processing, candidate models, and why to choose them. Follow-up: decide whether a two-percent price increase is viable and explain how to test or operate it.

Other reported topics or limits: Infer Windows Developer Feature Preferences Across Machine Types; Statistics, Regression, Ensemble, and Tree Fundamentals.

068: Data & Applied Science Intern — phone or technical screen

Report date: 2021-10-21 · Role: Data & Applied Science Intern · Stage: Phone / technical screen

Microsoft Data & Applied Science internship recruiter-screen report. The rapid call covered proudest and most challenging projects, motivation, preferred DS/ML area, data workflow, difficult teamwork, DS experience, NLP familiarity, programming language, a built deep-learning model, explaining deep learning to a child, and accommodating user needs. No later interview update was added.

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: Project, Data Workflow, NLP/DL, and Behavioral Screen.

069: Data/Applied Scientist Intern — phone or technical screen

Report date: 2021-10-13 · Role: Data/Applied Scientist Intern · Stage: Phone / technical screen · Interview date: 2021-10-13

Detailed Microsoft Data/Applied Scientist internship phone screen. The technical portion asked for verbal explanations of overfitting remedies, Ridge versus Lasso and variable selection, binary-classification metrics, and bagging versus boosting, followed by behavioral questions. No reusable implementation or design artifact was requested.

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: Overfitting, Regularization, Classification Metrics, and Ensemble Theory Screen.

070: Data & Applied Science PhD Intern — final interview loop

Report date: 2021-10-02 · Role: Data & Applied Science PhD Intern · Stage: Final loop / onsite · Interview date: 2021-09-23

Detailed Microsoft Data & Applied Science PhD internship offer report.

1. LeetCode 1662: Check If Two String Arrays Are Equivalent. Determine whether two lists of string fragments represent the same word when concatenated. Example: ['Mic','ro','soft'] and ['M','ic','roso','ft'] both represent Microsoft. The candidate could use any language and run the code.

Other reported topics or limits: Diagnose a Metric Drop, Build the Dataset, and Choose a Classification Threshold; Unnamed Simple SQL and Statistics/ML Fundamentals.

071: Data & Applied Science Intern — final interview loop

Report date: 2021-09-30 · Role: Data & Applied Science Intern · Stage: Final loop / onsite · Interview date: 2021-09-23

Microsoft Data & Applied Science summer-intern offer report.

1. Evaluate Monthly Active Device Estimates Without Ground Truth. A model estimates monthly active devices. There is no direct ground-truth label for the target metric. Design how to evaluate the model's predictions.

Other reported topics or limits: Identify Required Data and Write a Hypothetical SQL Query; Naive Bayes Spam Explanation and Behavioral Questions.

072: Data Scientist PhD Intern — phone or technical screen

Report date: 2021-09-30 · Role: Data Scientist PhD Intern · Stage: Phone / technical screen

Microsoft Data Scientist PhD internship first-round screen. The candidate received behavioral and background questions plus a child-friendly explanation of deep learning. There was no coding. The candidate later confirmed that the second round was the final round but did not share its questions.

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: Behavioral, Research, and Child-Friendly Deep Learning Screen.

073: Data & Applied Science PhD Intern — final interview loop

Report date: 2021-09-29 · Role: Data & Applied Science PhD Intern · Stage: Final loop / onsite · Interview date: 2021-09-23

Detailed Microsoft Data & Applied Science PhD internship offer report. The initial screen covered broad background and ML communication questions. Four onsite rounds were team-specific case studies requiring end-to-end analysis, data acquisition, modeling, and evaluation, plus one undisclosed SQL task. The candidate confirmed there was no coding.

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: Four Undisclosed Team-Project Case Studies; One Undisclosed SQL Task; Background, ML Communication, and Product Opinion Screen.

074: Data & Applied Scientist Intern — phone or technical screen

Report date: 2021-09-29 · Role: Data & Applied Scientist Intern · Stage: Phone / technical screen · Interview date: 2021-09-28

Microsoft Data & Applied Scientist Intern first-round report from a third-year PhD. The 30-minute Teams interview contained ten minutes of behavioral questions and a rapid ML-concept round with no coding.

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: Rapid ML Concepts and Behavioral Screen.

Part 7 continues with the next candidate reports.