The Walt Disney Company · Phone screen → Onsite
In processDisney Experienced MLE Interviews: 3 Experiences (Part 14 of 18)
Disney experienced MLE interviews in 2025: Two Sum variants, vector ranking and SQL, then recommendation serving, feature stores and ML deployment cases.
At a glance
- Process: recruiter → phone screen (coding, ML or SQL) → hiring manager → onsite with coding, ML system design and ML case rounds (entries 01–03).
- Asked recently: Two Sum, Three Sum and sorted variants (01); online recommendation serving (01); ranking vectors by similarity and text classification (02); a feature-store upsert pipeline and inference-cost and model-release cases (03).
- Difficulty: coding stayed at LeetCode easy to medium; the senior candidate got a verbal offer (03).
- Practise first: Rank vectors by similarity, feature store design and Three Sum.
Coverage: 3 experienced machine learning engineer reports from April to August 2025 (one senior).
In this series
New Grad SWE / SDE
- Part 1 — NEW GRAD · SWE/SDE: 2026 hiring context; 2025-cohort return offer; older interviews.
- Part 2 — NEW GRAD · SDE1: 2022 OA, coding and hiring updates.
- Part 3 — NEW GRAD: SDE1 August 2022 back to 2021; fullstack kept separate.
- Part 4 — NEW GRAD · SDE/SDE1: 2021 graduate interviews, chronology continued.
Other New Grad roles
- Part 5 — NEW GRAD · OTHER ROLES: MLE assessment in 2025; scientist interview in 2019.
Hiring cohort unconfirmed
- Part 6 — COHORT UNCONFIRMED: 2026 SWE I reports, other SWE and DE accounts.
- Part 7 — COHORT UNCONFIRMED: SDE1, OA and Android accounts from 2021–2022.
Experienced Full-Time
- Part 8 — EXPERIENCED · SWE: 2026 Staff/Product and 2025 backend/Ads/Principal.
- Part 9 — EXPERIENCED · SWE: 2026 array Set; 2025 ad filtering, backend and design.
- Part 10 — EXPERIENCED: frontend 2025, then a separate fullstack section for 2024.
- Part 11 — EXPERIENCED · SWE HISTORY: 2024 coding/Ads/Principal, then 2018 BQ.
- Part 12 — EXPERIENCED · SDE1: 2021–2022 early-career hires, not graduate hiring.
- Part 13 — EXPERIENCED · DATA ENGINEER: 2026 cloud and hybrid-pipeline interviews.
- Part 14 (current) — EXPERIENCED · MLE: 2025 coding, SQL, feature stores and modeling.
- Part 15 — EXPERIENCED · OTHER ROLES: GenAI 2025, QA 2024 and data-team 2020.
- Part 16 — EXPERIENCED · BAMTECH: separate software and Senior DE screens in 2019.
Contract
- Part 17 — CONTRACT · DATA ENGINEER: 2025–2026, separate from full-time hiring.
Internship / Co-op
- Part 18 — INTERNSHIP: software/PPO, AI Product and Technical Artist; roles separated.
Machine learning engineering
What came up
These 3 candidate accounts share the role/cohort scope stated above. Counts apply to this section only.
| Topic | Reports | Last seen |
|---|---|---|
| ML fundamentals / projects | 3 | Aug 2025 |
| Coding / data structures | 2 | Aug 2025 |
| System design | 2 | Aug 2025 |
| MLOps | 1 | Jun 2025 |
| ML modeling / evaluation | 1 | Apr 2025 |
| SQL / data | 1 | Apr 2025 |
| Behavioral questions | 1 | Apr 2025 |
Questions we found
Algorithms / Data Structures
| Question | Reports | Last seen |
|---|---|---|
| Two Sum | 1 | 2025 Q3 |
| Sorted Two Sum returning all unique index pairs | 1 | 2025 Q3 |
| Three Sum | 1 | 2025 Q3 |
| Rank vectors by similarity to a query vector | 1 | 2025 Q2 |
System Design
| Question | Reports | Last seen |
|---|---|---|
| Online recommendation model serving | 1 | 2025 Q3 |
| Batch upsert pipeline from S3 into a feature store | 1 | Apr 2025 |
| Dashboard for displaying inference outputs | 1 | Apr 2025 |
SQL / Data
| Question | Reports | Last seen |
|---|---|---|
| Participants with total time >30 minutes and every attempt within 10 minutes | 1 | Apr 2025 |
| Count buggy and bug-free code submissions per employee | 1 | Apr 2025 |
ML Modeling
| Question | Reports | Last seen |
|---|---|---|
| Text classification for posts and email | 1 | 2025 Q2 |
| Viewing-device feature representation and disappearing data | 1 | Apr 2025 |
| Validate a new inference model with limited business risk | 1 | Apr 2025 |
| Reduce inference cost for a model with over 2000 features | 1 | Apr 2025 |
Candidate timelines
01: MLE: Two Sum variants and recommendation serving
Interview period: 2025 Q3 (as reported) · Report: 2025-08-22 · Role: Machine learning engineer · Cohort: Experienced Full-Time
The candidate was interviewing after a July 2025 layoff. The recruiter screen covered ML fundamentals. Technical coding went from Two Sum to sorted Two Sum returning all unique index pairs, then Three Sum.
The onsite combined behavioral discussion, online recommendation-model serving design, ML project discussion and fundamentals with follow-ups. The source marks the process as passed; it does not separately establish an accepted offer. The exact serving workload and coding boundary rules were not recorded in full.
02: MLE phone screen: classification, vector ranking and ML operations
Interview period: 2025 Q2 (as reported) · Report: 2025-06-06 · Role: Machine learning engineer · Cohort: Experienced Full-Time
The June 2025 phone screen came through a referral and lasted about an hour. Five areas were covered. First came text classification for posts or email: tokenization, feature construction and model choice. There was no supplied dataset or fully specified modeling assignment, and no verified Disney practice destination for this discussion.
Coding asked the candidate to rank other vectors by similarity to a query vector. The source does not name the similarity metric. The linked exercise uses dot product as an explicit practice assumption, rather than claiming the interviewer required it.
The other areas were an LLM resume project, neural-network versus Transformer fundamentals, and ML operations. A hiring manager shadowed without asking questions. The candidate received a generic rejection two days later.
03: Senior MLE: linked HM and onsite reports, then a verbal offer
Interview period: 2025 Q2 (as reported) · Report: 2025-04-15 · Role: Senior machine learning engineer · Cohort: Experienced Full-Time
The candidate explicitly linked the hiring-manager account and the later onsite report, so they form one interview cycle here. The one-hour HM interview was about a month before the April 14, 2025 report and had resume/BQ, ML breadth, and data-engineering/BI sections. It covered resume and project questions, feature engineering, model selection, production deployment, bias/variance, precision/recall calculations, ROC/AUC use cases, and a conditional-probability cancer-detection question. The discussion also included a dashboard for inference outputs, large-table merge/upsert optimization, and ETL experience with AWS Glue, Lambda, Data Wrangler and Athena. Passing this HM conversation was not yet an onsite pass.
An onsite invitation arrived a week after the HM screen. The four onsite sessions ran over three days, ending roughly a week before the April 15 report; coding and system-design work used HackerRank:
- After introductions, BQ and resume/project discussion, a senior data scientist discussed viewing-device features: per-device percentages versus raw viewing hours for one account, granularity, table design and what to do when data for a device category disappears. The planned half-hour session ran over and left little time for the candidate's questions.
- In a one-hour round, a senior MLE gave SQL over participant (ID, name, group) and submission (submission ID, start/end timestamps, verdict) tables: total time above 30 minutes while every individual attempt was completed within 10 minutes, sorted by total time ascending, followed by window-function questions. The discussion then covered data-lake/mart layers, an AWS feature store, embeddings and prompt tuning.
- A lead MLE discussed prior modeling, features, metrics, PySpark, SageMaker, MLflow and an LLM chatbot project. Despite the recruiter's description of Python coding, this hour was project discussion rather than a reported Python exercise.
- A second one-hour lead-MLE round used employee (ID, name, department) and code-submission (submission ID, employee ID, timestamp, bug flag) tables to count buggy and bug-free submissions per employee, ordered by name. Design covered batch upserts from S3 through cleaned and aggregated layers into a feature store, including the purpose of each layer, ingestion/processing between layers, where to run quality checks, and how to monitor the resulting feature quality. The candidate was asked how to validate inference changes before production and measure A/B business impact, and how to reduce the cost of a model with more than 2,000 features.
The candidate initially feared rejection while waiting. A May 14 follow-up says a verbal offer arrived after roughly two weeks. The April 14/15 dates are report dates, not exact dates for every round.
Common questions
What coding does Disney ask MLE candidates?
Two Sum and Three Sum variants (01), ranking vectors by similarity to a query (02), and SQL on participant times and code submissions (03).
What ML design does Disney ask?
Online recommendation model serving (01), a batch upsert pipeline from S3 into a feature store, an inference dashboard, and cases on validating a new model and cutting inference cost (03).
Keep practicing
- From these reports: the coding, SQL, design and ML cases are linked in entries 01–03, including Two Sum, Three Sum, vector ranking, recommendation serving and feature store.
- Disney question banks: algorithms and system design (sign-in needed).
- Everything Disney in one place: the Disney interview page.
- More firsthand reports: the Experience feed.
In this series
- Part 1
- Part 2
- Part 3
- Part 4
- Part 5
- Part 6
- Part 7
- Part 8
- Part 9
- Part 10
- Part 11
- Part 12
- Part 13
- Part 14 (current)
- Part 15
- Part 16
- Part 17
- Part 18
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.