TikTok Machine Learning Engineer Interview Questions
& Process
Real candidates share what happened, how many rounds they had,
and how the experience turned out.
Based on 236 interview experiences · FREE TO READ
Candidate interview experiences
First-hand accounts from people who interviewed at TikTok.
TikTok
Machine Learning Engineer
Mid Average Positive experience Accept offer 3 rounds
Interview process
Technical screen Technical screen Technical screen
Interview formats
Coding Technical
It was a 3 round interview process. About half the time was spent on LeetCode problems, and the other half on ML basics. They asked a binary search question and a graph traversal problem, both medium level LeetCode. For ML basics, the questions covered LLMs, attention, basic model structures, normalization techniques, and how to prevent overfitting.
Confirmed questions1 question
- Regarding LLMs, what kind of normalization is employed and what's the reasoning behind it?
TikTok
Machine Learning Engineer Intern
Intern Average Neutral experience No offer 1 round
Interview process
Technical screen
Interview formats
Coding Technical
I had a 60-minute first-round interview. It started with a 15-minute deep dive into my resume. Then, there were 20 minutes for a LeetCode-style coding problem and another 20 minutes for ML concepts. It ended with a Q&A. The interviewer was really nice and even gave me some hints when I was struggling with the coding part. Even though I didn't quite make it in the end, I still managed to get to the second round.
Confirmed questions3 questions
- Can you discuss your previous projects and highlight your most accomplished work?
- Could you solve a medium-difficulty coding problem that involves a sliding window approach?
- What is the time complexity associated with attention mechanisms?
TikTok
Machine Learning Engineer
Mid Average Neutral experience Decline offer 6 rounds
Interview process
Recruiter call Phone screen Technical screen Onsite Background check Offer
Interview formats
Coding Technical System Design
Got a recruiter reach out first, then a convo with the recruiter. They wanted to make sure I was a good fit for a team before any interviews, so I ended up waiting about 2 weeks to get matched. After that, there was a phone screen involving a LeetCode medium problem. The onsite portion was pretty extensive: two ML design rounds, two coding rounds, and a final round with the hiring manager. The whole thing took close to a month to complete.
Confirmed questions1 question
- Can you describe how you would model a signal and then leverage it to enhance an ads ranking system?
TikTok Machine Learning Engineer Interview Questions
Quoted word for word from TikTok interview reports.
“For N independent and identically distributed Gaussian random variables, what distribution does their linear combination follow?”
Read reports →“Can you find the sum of the two largest numbers that are still smaller than k, given an array of positive numbers that isn't sorted?”
Read reports →“Can you provide the formula for computing the attention score given Q, K, and V?”
Read reports →“Given the BytePlus platform with n servers, each storing data represented by the array memory, you can perform an operation any number of times: choose an index idx (1 ≤ idx ≤ n/2) and swap the data of servers equidistant from the beginning and end of the array memory, with a distance less than or equal to idx. The total working efficiency is the sum of the product of server data and its index. Given n and the array memory, find the maximum possible total working efficiency, printing the result modulo 10^9+7.”
Read reports →“Can you briefly describe what a transformer is, and what's the difference between a transformer and an RNN?”
Read report →“For predicting TikTok video CTR, how would you identify the most important features from a large, sparse set?”
Read report →“Can you introduce dropout and explain why BN works, and also why BN needs reshift and rescale?”
Read report →“Describe how to use binary search in a more complex data structure, like searching on the answer space.”
Read report →“Regarding optimizer usage in ML, can you compare SGD and Adam, and explain learning rate scheduling?”
Read report →Formats, difficulty and experience
Across all 236 TikTok interview reports.