ByteDance · Phone screen
RejectedByteDance USDS New Grad SWE Phone Screen: 1 Experience
A ByteDance (TikTok) USDS new grad SWE phone screen in late 2026 was one hour: a résumé project, a topological-sort coding variant and ML metric questions.
At a glance
- Process: one 60-minute technical phone screen on HackerRank after a referral: project walkthrough, coding, then spoken ML questions (report 01).
- Asked recently: a course-ordering variant solved with topological sort; metrics for extremely imbalanced data; precision and recall trade-offs (report 01, report date Nov 2026).
- Difficulty: rated average; the candidate was stuck on the coding and needed step-by-step help, and was rejected four hours later.
- Practise first: Lexicographically Smallest Topological Order and pick a project from your résumé.
Coverage: One new grad with a PhD, referred for a full-time software engineering role. The report's title names TikTok's USDS team and its company field says ByteDance, which is where it's listed here. It covers a one-hour technical phone screen, gives October to December 2026 as the interview period, and went up on November 11, 2026.
What came up
One hour held three parts: a project from the résumé, a graph-ordering coding question on HackerRank, and spoken machine learning questions about measuring models on very imbalanced data.
| Topic | Reports | Last seen |
|---|---|---|
| Coding: topological sort (a course-ordering variant) | 1 of 1 | Nov 2026 (report date) |
| ML foundations: metrics for imbalanced data, precision and recall | 1 of 1 | Nov 2026 (report date) |
| Résumé project walkthrough | 1 of 1 | Nov 2026 (report date) |
Questions we found
Algorithms / Data Structures
| Question | Reports | Last seen |
|---|---|---|
| A medium variant of a well-known course-ordering problem, solved with topological sort (Lexicographically Smallest Topological Order) | 1 | Nov 2026 (report date) |
ML Knowledge
| Question | Reports | Last seen |
|---|---|---|
| Which metric would you use when the data is extremely imbalanced? | 1 | Nov 2026 (report date) |
| Explain precision and recall from the confusion matrix, and how the right trade-off changes between settings such as medicine, finance and fake-account risk work | 1 | Nov 2026 (report date) |
Behavioral (BQ)
| Question | Reports | Last seen |
|---|---|---|
| Pick a project from your résumé and walk through it, then take follow-up questions (practice) | 1 | Nov 2026 (report date) |
1 detailed candidate timeline
01: USDS new grad phone screen, one hour in three parts
Interview period: 2026 Q4 · Report date: November 11, 2026 · Role: Software engineer, new grad · Degree listed: PhD · Channel: Referral · Candidate's rating: Positive, average difficulty · Result: Not selected
Technical phone screen (one hour, HackerRank)
- The interviewer introduced themselves, then the candidate did the same.
- The candidate picked a project from their résumé, explained it, and answered the interviewer's follow-up questions (practice).
- Coding: a medium-level variant of a well-known course-ordering problem, which the candidate identified by its problem number. It is solved with topological sort, and the candidate notes it can be approached from either direction. They were stuck for a long time. The interviewer guided them step by step and even pasted a topological sort into the editor. No tests were run at the end; the interviewer read the code and accepted it. The report doesn't say what the variant changes, so FastPrep's version, Lexicographically Smallest Topological Order, returns the smallest valid order as a practice choice.
- ML foundations, with no coding. First: which measure to use when the data is extremely imbalanced. Hints pointed the candidate to AUC and then to PR-AUC. Then: precision and recall from the confusion matrix. The discussion turned to how medical and finance settings care about different errors than the team's risk work on fake accounts. The candidate felt they got tangled here and didn't explain it clearly.
How it went
The candidate describes both the interviewer and HR as kind. The interviewer kept guiding and giving hints. HR had reached out to set up the interview and moved the process along quickly.
Result
The interview ended at 5 p.m., and a rejection arrived around 9 p.m. the same evening.
Common questions
What is the ByteDance USDS new grad phone screen?
One hour in this report: a résumé project walkthrough, a topological-sort coding variant on HackerRank, and spoken ML questions on imbalanced-data metrics and precision versus recall (report 01).
How fast does ByteDance reply after the phone screen?
About four hours: the interview ended at 5 p.m. and the rejection came around 9 p.m. (report 01).
Keep practicing
- Coding: Lexicographically Smallest Topological Order is the practice version of the phone screen's coding question. More ByteDance questions, in every practice format, are in the ByteDance question bank.
- Behavioral: the ByteDance behavioral question bank has the project walkthrough question linked above.
- Everything ByteDance in one place: the ByteDance interview page and ByteDance hiring insights.
- Other ByteDance cohort from this batch: a research scientist intern's first and second rounds on the Trust & Safety team are in this intern Experience.
- More firsthand reports: the Experience feed.
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.