RepostedStripe logo

Stripe · Phone screen

In process

Stripe Experienced MLE Phone Screens: 1 Experience

Machine learning engineer · Full-timeMaster's degreeReported Nov 5, 2026
Shared byjanebeingjane

A Stripe experienced MLE passed two phone screens in late 2026: account record linkage in coding, then an ML round focused on data more than the model.

At a glance

Coverage: One candidate changing jobs, interviewing for a full-time machine learning engineer role at Stripe after applying online. It was their first time interviewing for an MLE role. The report covers two technical phone screens, both passed, with an onsite still to come. It gives October–December 2026 as the interview period and went up on November 5, 2026.

What came up

The coding round was about linking similar accounts, first by direct matches and then through chains of matches. In the ML Integration round, the follow-ups were about data handling and model evaluation, more than about the model.

Questions we found

Algorithms / Data Structures

Question Reports Last seen
Record matching: decide which accounts are the same person from their metadata (Record Linkage Part 1 - Directly Linked Records) 1 Nov 2026 (report date)
Follow-up: matches chain together, so find the connected group (Record Linkage Part 3 - Full Connected Component) 1 Nov 2026 (report date)

ML Integration

Question Reports Last seen
An end-to-end modeling exercise on a Kaggle-style dataset, probed on data handling (duplicate keys, missing values, one-hot encoding) and on evaluation (recall, precision, AUC, calibration); the report doesn't describe the dataset or target 1 Nov 2026 (report date)

1 detailed candidate timeline

01: Account linking, then an ML round focused more on data than on the model

Interview period: 2026 Q4 · Report date: November 5, 2026 · Role: Machine learning engineer, full-time · Degree listed: Master's · Channel: Applied online · Candidate's rating: Positive, average difficulty · Result: Passed both phone screens; onsite next

Of everyone who has interviewed them, the candidate rates these two interviewers the best. Both showed a lot of patience and stepped in to help, with hints along the way and some scaffolding code so the tests ran sooner.

Round 1: Coding

A record-matching problem over user accounts that each carry some metadata.

  1. The opening parts were fairly basic: two accounts count as the same person when their fields agree, the name being one example, and the work is checking accounts against each other pair by pair. You can practice this as Record Linkage Part 1 - Directly Linked Records.
  2. Part 3 makes matching transitive: accounts linked through a chain of matches belong together even when the two ends don't match each other directly. The candidate treats it as a connected-components problem, solvable with BFS or union-find. FastPrep's version is Record Linkage Part 3 - Full Connected Component.

Both practice versions score field matches with per-field weights against a threshold and start from one target record: Part 1 returns the records linked to it directly, and Part 3 returns the rest of its connected group. This report only says that matching fields mean the same person.

Round 2: ML Integration

The setup is a dataset in the style of a Kaggle competition, taken from raw data to a finished model. The candidate notes that earlier reports say little about this round and guesses that it draws on a single question.

  • What the interviewer cared about, first: the data. Exploration and preprocessing, missing values and how to fill them, outliers, whether a DataFrame has duplicates on one key or a pair of keys, and one-hot encoding for categorical columns, including what happens to missing values there.
  • Then: evaluation, once a model existed. Recall, precision and AUC came up, followed by a detailed discussion of calibration and how to carry it out.
  • What mattered less: the model itself. It wasn't the focus, and a random forest was enough.

For this round, the candidate suggests getting comfortable with Kaggle-style problems.


Other Stripe cohort: a new-grad software engineering report is in 2026 Stripe New Grad SWE Phone Screen, Onsite and HM Round: 1 Experience.


Common questions

What is the Stripe MLE coding round?

In this report, a record-linkage problem: decide which accounts are the same person from their metadata, then group accounts linked through chains of matches (report 01).

What does the Stripe ML integration round test?

Data handling first (exploration, missing values, duplicate keys, one-hot encoding), then evaluation (recall, precision, AUC, calibration); the model itself mattered less (report 01).

Keep practicing

Sources: Anywhere useful interview info can be found.

0 total reactions. Your reactions: none.
0 comments

Comments

0

No comments yet — be the first to share what you know.