Adobe ML and Research Intern Round 1: 4 Experiences (Part 2 of 2)
- Company
- Adobe
- Reports
- 4 interview reports
- Level
- Internship
- Stages
- Phone screen
Adobe ML, research and web developer intern round 1 reports from 2024–2026 center on the candidate's own projects, with puzzles for research roles.
At a glance
- Process: an online assessment came first for three of the four (15, 16, 18) and a recruiter screen for the fourth (17). The interview ran 40–45 minutes (16), about an hour (17) or 75 minutes with a manager and a teammate (15).
- Asked recently: project or research deep dives in 3 of 4 (15, 16, 17), puzzles or probability in both research rounds (16, 17), and live coding in 2: Top K N-gram Anagram Groups (15) and Group Words by Their Letters (18).
- Difficulty: candidate 15 got hints on the coding and was told they did a great job; 16 needed a hint on the brick puzzle. Three of the four got offers (15, 16, 18), and 18 declined.
- Practise first: Top K N-gram Anagram Groups, Group Words by Their Letters and explain a confusion matrix to someone without a technical background.
Coverage: 4 intern reports from Nov 2024 to Mar 2026: a Summer 2026 machine learning intern (U.S.), an Adobe Research intern (India, on campus), a machine learning research intern (U.S.) and a web developer intern (Bucharest). First live technical round only. Each role has its own section below.
In this series
- Part 1: software engineering interns — 14 reports, entries 01–14, with the overview
- Part 2 (current): ML, research and web developer interns, entries 15–18
Machine learning intern: what came up
The one ML intern round paired live coding (top-k anagram groups of word n-grams) with project, ML and teamwork questions.
Machine learning intern: questions we found
Algorithms / Data Structures
| Question | Reports | Last seen |
|---|---|---|
| Top K N-gram Anagram Groups, with scale, audio-input and evaluation follow-ups | 1: 15 | Mar 2026 |
Machine Learning
| Question | Reports | Last seen |
|---|---|---|
| Fine-tuned LLM project: how it was evaluated, why that model, how the data was prepared | 1: 15 | Mar 2026 |
Behavioral (BQ)
| Question | Reports | Last seen |
|---|---|---|
| Explain a confusion matrix to someone without a technical background (practice) | 1: 15 | Mar 2026 |
| Two developers propose different solutions to one problem. How do you handle it? (practice) | 1: 15 | Mar 2026 |
Machine learning intern: 1 detailed candidate timeline
15: Summer 2026 ML intern, n-gram top-k coding plus project questions
Interview date: Mar 16, 2026 (two days before the offer) · Role: Machine Learning Intern (Summer 2026) · Location: United States · Result: Offer
Before the first technical round: An online assessment with three Python questions: an easy anagram-grouping task and two Python OOP questions.
75-minute interview with a manager and a teammate (first technical round)
- Project questions about a fine-tuned LLM on the resume: how it was evaluated, why that model was chosen, and how the dataset was prepared. They also asked the candidate to explain a confusion matrix to someone without a technical background.
- Live coding on HackerRank: split text into chunks, build word n-grams, group the ones that are anagrams, and return the top-k groups by frequency. The candidate wrote, explained and tested the code. The panel gave hints along the way.
- Follow-ups on the coding task: how to scale it to millions of records, how to handle audio input instead of text, and how to evaluate the top-k output. The manager told the candidate they did a great job.
- A behavioral question: two developers propose different solutions to the same problem. How do you handle it?
Research interns: what came up
Both research intern rounds started with the candidate's own research, and both added puzzle or probability questions.
| Topic | Reports | Last seen |
|---|---|---|
| Puzzles and probability | 2 of 2 | Jul–Aug 2025 |
| Resume and project deep dive | 2 of 2 | Jul–Aug 2025 |
| Machine learning | 1 of 2 | Nov 2024 |
Research interns: questions we found
Behavioral (BQ)
| Question | Reports | Last seen |
|---|---|---|
| What is new about your research project, and how does it beat existing methods? (practice) | 1: 17 | Nov 2024 |
Puzzles and Probability
| Question | Reports | Last seen |
|---|---|---|
| Two ants on a 1 m string moving at 1 m/s: where are they after one second? | 1: 16 | Jul–Aug 2025 |
| Can 1×1×4 bricks fill a 6×6×6 cube, and why not 53 of them? | 1: 16 | Jul–Aug 2025 |
Research interns: 2 detailed candidate timelines
16: Adobe Research intern on campus, two puzzles and research talk
Interview date: Jul–Aug 2025 (a few weeks before the Aug 19, 2025 report) · Role: Research Intern (Adobe Research) · Location: India (campus) · Result: Selected the same night
Before the first technical round: A 90-minute online assessment: 20 multiple-choice questions on logic, probability, statistics, ML, deep learning and LLMs, plus two medium coding problems (a sweep line and a DP on arrays).
Technical interview (first technical round)
- The interview was the next day and lasted 40 to 45 minutes. It started 25 to 30 minutes late, and the interviewer didn't introduce themselves.
- Questions on the candidate's current research.
- Puzzle: two ants sit at random spots on a 1 m string and each moves at 1 m/s in either direction. Where are they after one second? When the candidate asked, the interviewer said collisions are elastic. The candidate answered that each ant ends at its start point d or at the mirrored point 1 − d, and the interviewer was satisfied.
- Puzzle: can 1×1×4 bricks fill a 6×6×6 cube, and why can't 53 of them fit? The candidate started with a volume argument. The interviewer hinted at the chessboard with two opposite corners removed, and the candidate moved to a coloring argument.
- A few quick probability questions.
- Resume: research projects, past internships and a hackathon contribution.
17: ML research intern, an hour with a team scientist
Interview date: Nov 2024 · Role: Machine Learning Research Intern · Location: United States · Result: No offer
Before the first technical round: A recruiter phone screen.
Technical interview with a team scientist (first technical round)
- About an hour long. The first 30 to 35 minutes were a deep dive into a research project on the resume: what is new about it, and how it beats existing methods.
- The last 20 minutes were basic computer vision and probability questions. The exact questions weren't shared.
Web developer intern: what came up
The one web developer intern round was JavaScript from start to finish: grouping anagrams, array methods, hoisting, HTTP status codes and React hooks.
Web developer intern: questions we found
Algorithms / Data Structures
| Question | Reports | Last seen |
|---|---|---|
| Group Words by Their Letters (LeetCode 49), in JavaScript | 1: 18 | Nov 2024 |
Frontend / UI Coding
| Question | Reports | Last seen |
|---|---|---|
| React hooks and state management | 1: 18 | Nov 2024 |
Language and CS Fundamentals
| Question | Reports | Last seen |
|---|---|---|
| JavaScript: map, filter, reduce and currying | 1: 18 | Nov 2024 |
| JavaScript: arrow functions, destructuring and hoisting | 1: 18 | Nov 2024 |
| REST and HTTP status codes | 1: 18 | Nov 2024 |
Web developer intern: 1 detailed candidate timeline
18: Web developer intern in Bucharest, anagrams in JavaScript and React hooks
Interview date: Nov 2024 · Role: Intern Web Developer · Location: Bucharest, Romania · Result: Offer, declined
Before the first technical round: An online assessment.
Online technical interview (first technical round)
- A JavaScript coding task: group words that are anagrams of each other (LC 49), for example "abc", "bca" and "cab" in one group.
- What map, filter and reduce do, and what currying is.
- Arrow functions, destructuring and variable hoisting.
- REST and HTTP status codes.
- How React hooks work and how to manage state.
After round 1: An in-person soft-skills interview about teamwork and agile.
Patterns across the reports
- Research first. Both rounds opened with the candidate's own research, then moved to puzzles or probability.
Common questions
Do Adobe research intern interviews include puzzles?
Both research rounds did. Candidate 16 had two puzzles (ants on a string, bricks in a cube) and some quick probability; 17's last 20 minutes were computer vision and probability questions.
What coding does the Adobe ML intern round ask?
The one ML intern report had live coding on HackerRank: build word n-grams, group the anagrams and return the top-k groups, with follow-ups on scale, audio input and evaluation (15).
Keep practicing
- Top K N-gram Anagram Groups: practice problem from report 15.
- Two teammates each propose their own way to solve one problem. How do you settle on what to do?: behavioral question from report 15.
- A colleague without a technical background asks what a confusion matrix shows. How do you explain it?: behavioral question from report 15.
- What does your research project add that earlier work did not, and why does it beat earlier approaches?: behavioral question from report 17.
- Group Words by Their Letters: practice problem from report 18.
- Adobe coding questions: the company-filtered question bank (sign-in needed).
- Adobe behavioral questions: the company-filtered question bank (sign-in needed).
- Adobe ML modeling questions: the company-filtered question bank (sign-in needed).
- Adobe interview guide: Adobe's interview overview page on FastPrep.
- Adobe hiring insights: hiring activity and trends for Adobe.
- Experience feed: more interview experiences and roundups.
In this series
- Part 1
- Part 2 (current)
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.