TikTok Interview Questions — Page 5

TikTok·Data Scientist·Technical Phone Screen

Apr 2026

TikTok DS interview focused entirely on a single open-ended experiment design case about recommendation diversity. Three connected parts plus follow-ups, all building on the same scenario. Felt more like a product-DS hybrid than a pure stats round.

You're launching an A/B test for a new recommendation algorithm meant to increase content exploration. Walk through how you'd design and run that experiment to decide whether to launch.
How would you define metrics for exploration, long-term user satisfaction, and creator ecosystem health in this context?
The experiment window is too short to observe long-term impact. How do you analyze the data you have and make a launch recommendation anyway?
Exploration metrics go up but like rate drops. Do you ship?
How do you stop the model from just surfacing random irrelevant content in the name of diversity?
How would you measure whether users actually return to creators they discovered through the new algorithm?

“This is where I spent most of my time and honestly fumbled the randomization discussion a bit.”

View Post

TikTok·Product Manager·Onsite - Product Sense / Strategy

Apr 2026

Interviewed for a PM role at TikTok and got hit with a product metrics question framed around Instagram, which was a bit of a curveball given the company.

How would you measure whether Instagram's search feature is successful?

“Kind of a weird one to get at TikTok of all places.”

View Post

TikTok·Data Scientist·Technical Phone Screen

Apr 2026

TikTok data scientist interview focused heavily on classifier evaluation, specifically the precision-recall tradeoff and how you'd actually set a decision threshold in a real product context. Pretty applied, less about theory and more about whether you understand the business consequences of getting it wrong.

In the context of a binary classifier that triggers some product action, what do precision and recall actually mean from a business standpoint?
Walk me through two or three scenarios where you'd care more about precision than recall.
What are some scenarios where recall should take priority over precision?
How would you actually choose an operating threshold for a classifier in practice? Think through cost asymmetry, review capacity, class imbalance, what curves you'd use, and what you'd monitor post-launch.

“I started with the textbook definitions and then tried to map them to something concrete like content moderation.”

View Post

TikTok·Machine Learning Engineer·Onsite - System Design / Architecture

Apr 2026

TikTok MLE interview for a matching/ranking team. One open-ended ML system design question that I fumbled pretty badly, mostly because retrieval systems aren't really my area.

In a recommender system's recall stage, K candidates are retrieved and passed to a heavy ranker. Design a model or system that determines K dynamically rather than using a fixed value.

“I did not have a good answer for this and kind of knew it in the moment.”

View Post

TikTok·Machine Learning Engineer·Technical Phone Screen

Apr 2026

TikTok ML Engineer interview, focused entirely on machine learning fundamentals. Pretty theory-heavy, covering everything from basic overfitting to RL-based alignment for LLMs. No coding, just concepts, which I wasn't fully expecting.

What is overfitting and how do you mitigate it in machine learning?
In the context of deep learning specifically, what are common techniques to reduce overfitting?
Explain how dropout works during training, why it counts as regularization, and how you handle it at inference time.
Compare batch normalization and layer normalization: what statistics do each use, how do they behave differently across batch sizes and sequence models, and which parameters are actually used at deployment?
What are the common ways reinforcement learning is applied in LLM post-training for alignment or fine-tuning?

“Straightforward enough to start.”

View Post

TikTok·Software Engineer·Hiring Manager Screen

May 2026

Went through a screening round for a software engineer role at TikTok that was pretty heavily focused on what I'd been up to recently and how I talk about my own work. Less algorithmic than I expected, more of a 'convince me you actually did the thing' vibe.

Walk me through what you've been doing since leaving your last job. Are you currently working somewhere full-time, and if not, what's the reason?
Tell me about your most recent project: what was the scope, what did you personally own, what technical calls did you make, and what were the actual results?
What did you personally deliver on that project versus what the team handled?
What did you learn from that project, and what would you change if you did it over?

“I had a gap and knew this was coming, still fumbled the opening a bit.”

View Post

TikTok·Data Scientist·Technical Phone Screen

May 2026

TikTok data scientist interview that went deep into experiment design for a tipping feature, sequential testing, and multiple comparisons. Pretty technical throughout, felt like a stats exam more than a product conversation.

You have a creator-level randomized experiment with 10,000 creators per arm, each creator averaging 100 viewer sessions, a viewer-level purchase rate of 5% in control and 5.2% in treatment, and an intra-cluster correlation of 0.02. Compute the design effect and effective sample size, then derive a z-statistic and two-sided p-value using cluster-robust standard errors.
If you run 4 interim looks plus a final analysis on this experiment, how would you approximate an O'Brien-Fleming alpha-spending approach to control overall alpha at 0.05, and how does that compare to a naive Bonferroni correction? What are the implications for power and experiment duration?
With four guardrail metrics being monitored alongside the primary metric, walk through how you'd apply a Holm-Bonferroni correction, and describe when you'd switch to reporting Bayesian posterior intervals with a region of practical equivalence instead.

“This one took me a minute to set up correctly.”

View Post

TikTok·Machine Learning Engineer·Technical Phone Screen

May 2026

TikTok ML engineer interview that leaned hard into transformer internals and sampling methods. Two meaty coding questions back to back, both with follow-ups that required actual understanding rather than just pattern matching. Left feeling like I'd studied the right stuff but maybe not deeply enough.

Implement multi-head attention from scratch given query, key, and value tensors, the number of heads, and an optional mask. Your output should match the input sequence shape and include linear projections, scaled dot-product attention with softmax, masking, dropout, head concatenation, and a final output projection. Follow-up: why do we scale by the square root of the head dimension?
Implement nucleus (top-p) sampling for a language model: given a vocabulary's logit scores and a threshold p between 0 and 1, find the smallest set of tokens whose cumulative probability meets or exceeds p, renormalize over that set, and sample the next token. Follow-up: how does this compare to top-k sampling in terms of advantages and disadvantages?

“I got the structure right but fumbled the reshape logic for splitting heads.”

View Post

TikTok·Data Scientist·Onsite - Behavioral / Leadership

May 2026

TikTok DS interview with a single deep behavioral question about creator growth. The whole thing was basically one long prompt asking you to reconstruct an entire project lifecycle from scratch, which felt more like a case study than a standard behavioral round.

Walk me through a time you owned an ambiguous creator growth problem end-to-end. Cover the business goal, how you defined success and handled trade-offs between creator and viewer outcomes, a disagreement with a cross-functional partner, a decision you got wrong and the signals you missed, how you adapted for regional differences, and what you did to protect long-term platform health.

“This is a beast of a question.”

View Post

TikTok·Software Engineer·Technical Phone Screen

May 2026

TikTok software engineer interview with two JavaScript-heavy coding problems back to back. Nothing too exotic but the Promise.all one took more thought than I expected.

Write a function that flattens a nested JavaScript object into a single-level object where nested keys are represented as dot-separated paths (e.g., {a: {b: 1}} becomes {'a.b': 1}).
Implement Promise.all from scratch. It should accept an iterable of promises or plain values and return a promise that resolves with all results in order, or rejects immediately if any input rejects.

“Pretty standard once you see it a few times.”

View Post