Company interview guide
Tencent Machine Learning Interview Guide (2026)
Resume-driven technical rounds mixing ML fundamentals, a medium-level coding problem and deep questions on your own projects.
Tencent's machine learning interview loop has 7 reported stages, from online written test to hr interview; typical length: About 2 weeks to 2 months.
Roles covered: ML Engineer, Research Scientist and Data Scientist. The question bank holds 2 questions reported from Tencent interviews, 1 of them free to solve in your browser.
- Interview process length
- About 2 weeks to 2 months
- Stages
- 7
- Reports linked to bank questions
- 2
- Bank report dates
- May 2023–Apr 2024
- Questions in bank
- 2 (1 free)
- Last reviewed
- October 2026
How the Tencent interview loop works
Application or employee referral, then for many roles an online written test of algorithm problems, sometimes with multiple-choice math, logic or situational questions. Next come two to four technical interviews by phone or video, each about an hour, which open with a self-introduction and move between your resume projects, ML fundamentals and a live coding problem; data science rounds often add a short business case tied to the group's product. One of the later rounds is usually with the group lead, a senior engineer or the hiring manager, and an HR interview on motivation, pay and availability closes the loop. Referred candidates sometimes skip the written test. Round counts vary by group, from two interviews to six technical rounds plus a manager round in one US loop. Reported timelines run from about two weeks to two months. This outline rests on 25 first-hand reports, most from before 2024, so ask your recruiter which steps your group uses.
The Tencent interview process, stage by stage
01Online written test
Online, before any interview; mostly algorithm problems, sometimes multiple-choice math, logic or situational sections
Many candidates, especially through campus channels, first take an online test. Most describe data structure and algorithm problems at easy to medium level; one research test mixed multiple-choice math and programming with one coding problem, and a 2025 data science test had two 30-minute timed sections of logic and workplace scenario questions. Referred candidates have sometimes skipped it.
What it tests
- Solving easy to medium algorithm problems correctly under time pressure
- Basic math and programming knowledge in multiple-choice form
- Logical reasoning and prioritising limited time across many questions
- Judgment on workplace scenarios
How to prepare
- Practise timed sets of medium array, string and dynamic programming problems.
- For multi-section tests, skip and return rather than stalling on one item; reports stress the time limit.
- Ask whether a referral lets you skip the test for your role.
An online written or coding test before interviews in 6 ML, research and data science reports from 2019-2025; only one is from 2024 onward, and one referred candidate skipped it.
02Resume and project deep dive
Part of every technical round, often opening it; about an hour per round by phone or video
The most consistent theme across reports: interviewers build their questions from your resume and push into each project in detail. Expect to justify why you picked a model, how you evaluated it, what the data looked like and how your publications or research fit together. Interviewers also tie questions to the group's own business.
What it tests
- A clear self-introduction and summary of your best project
- Why you chose a given model or method, and what alternatives you rejected
- How you picked and interpreted evaluation metrics
- The theory behind every technique listed on your resume
- Research contributions and papers, for research roles
How to prepare
- Treat everything on your resume as fair game, including methods you only used briefly.
- For each project, prepare the problem, data, model choice, metric, result and what you would change.
- Read about the group's products before the round; questions often connect your work to their business.
Resume or project deep dives in 17 of 25 ML, research and data science reports, 5 of them from 2024-2025; the most consistent element of the loop.
03ML and deep learning fundamentals
Question-and-answer inside the technical rounds, usually linked to your projects
Interviewers check that you know the classic algorithms well: regularization, logistic regression versus SVM, how XGBoost works and how it differs from LightGBM, random forest feature importance. Deep learning and NLP or vision roles add embeddings, graph networks and transformers, and a 2024 algorithm engineer loop asked about the Transformer and tuning LLMs with RL. Data science rounds can include statistics and causal inference.
What it tests
- Classic models in depth: logistic regression, SVM, random forest, gradient boosting
- Regularization and evaluation metrics
- Embeddings, graph networks and transformer architecture
- Awareness of LLM training, including RL-based tuning
- Statistics and causal inference for data science roles
How to prepare
- Be able to explain XGBoost from the objective to the split finding, and how LightGBM differs.
- Prepare short comparisons: L1 versus L2, logistic regression versus SVM, bagging versus boosting.
- For algorithm and NLP roles, know transformer blocks, shapes and cost, plus the basics of RLHF.
ML or deep learning knowledge questions in 14 ML, research and data science reports; 4 are from 2024-2025, and tree models and logistic regression recur most.
04Live coding
Usually one problem inside a technical round, medium difficulty (some reports say up to hard); in person or by video
Most technical rounds include a standard algorithm problem that you write and then walk through. Reported examples include an LRU cache, edit distance, generating all balanced bracket sequences and inserting into a sorted list. One data science loop ran a dedicated coding interview first, and another included a live hands-on data task.
What it tests
- Medium algorithm problems: dynamic programming, backtracking, hash maps and linked lists
- Designing a small data structure with O(1) operations
- Explaining your code and its complexity
- Hands-on data work for data science roles
How to prepare
- Drill dynamic programming and backtracking classics until you can write them without hesitation.
- Talk through the approach before coding; interviewers ask you to walk them through the result.
- Data science candidates should also expect SQL or a practical data task.
Free reported questions for this stage
A live coding problem inside interviews in 10 ML, research and data science reports, 3 from 2024-2025; difficulty is mostly medium, one says medium to hard.
05Business and ML case (data science and applied roles)
A short case inside a technical round, tied to the group's product such as games or ads
Data science candidates often get a small case: design an ML pipeline for a games business problem, segment players or recommend content, or propose a data solution and business metrics for a real scenario. These are discussions, not slide presentations, and three of the five reports are from 2024-2025.
What it tests
- Framing a vague business question as an ML or analytics problem
- Choosing features, models and success metrics
- Segmentation and recommendation thinking for games and content products
- Practical judgment on data quality and deployment
How to prepare
- Prepare a reusable outline: goal, metric, data, model, evaluation, rollout.
- Know the product of the group you are joining; gaming roles asked about player segmentation and monetisation.
Business or ML case questions in 5 data science reports, 3 from 2024-2025; ML engineer and research reports rarely mention them.
06Group lead or hiring manager round
One technical conversation with the group lead, a senior engineer or the hiring manager, often the second or a later technical round
Several candidates single out the round with the group lead or manager as the decisive one. It covers the same ground as earlier rounds, often with harder follow-up on your projects and sometimes another coding question, and checks fit with the team's work.
What it tests
- Depth and ownership in your past work
- Fit between your skills and the team's current problems
- Coding or technical questions at the lead's discretion
How to prepare
- Prepare questions about the team's roadmap; the lead is often the one deciding.
- Expect the hardest follow-up questions on your strongest project here.
A separate group lead, senior or hiring manager round in 6 ML, algorithm and data science reports from 2017-2022; none of the 2024-2025 reports mention it.
07HR interview
Final conversation with HR on motivation, behavior, pay and availability
The loop usually ends with HR. Reports mention behavioral questions that are not easy to answer, discussion of salary and start date, and punctuality being noticed. One written test included a scenario about being kept away from a group's core project.
What it tests
- Motivation for Tencent and the specific group
- Handling workplace frustration and teamwork
- Salary expectations and availability
How to prepare
- Prepare a calm, constructive answer for scenarios about recognition and access to core work.
- Have a clear view of salary range and start date before this round.
An HR or behavioral closing interview in 6 ML and data science reports, 2 from 2024; several others mention HR only as the contact for results.
Questions and themes reported most often
- Walk through a resume project and defend every modeling choice (17 reports, resume and project deep dive)
- How XGBoost works and how it compares with LightGBM or random forest (3 reports, ml and deep learning fundamentals)
- Logistic regression: why use it, and how it differs from SVM (3 reports, ml and deep learning fundamentals)
- ML pipeline, segmentation or recommendation case for a games or product business (3 reports, business and ml case (data science and applied roles))
- LRU cache with O(1) get and put (1 reports, live coding)
- Edit distance by dynamic programming (1 reports, live coding)
- Generate all balanced bracket sequences of length 2n (1 reports, live coding)
- Transformer architecture and tuning LLMs with RL (1 reports, ml and deep learning fundamentals)
What Tencent asks about
- Resume & project deep dive30%
- ML & DL fundamentals27%
- Coding (written test and live)25%
- Data & business cases9%
- HR & behavioral9%
Tencent interview prep checklist
- Can explain each resume project end to end, including model choice and metric
- Can explain XGBoost and compare it with LightGBM and random forest
- Can compare logistic regression with SVM and L1 with L2 regularization
- Solved LRU cache, edit distance and balanced parentheses without help
- Prepared a case outline for a product or games problem in the target group
- Have questions ready for the group lead about the team's work
- Know your salary expectations and availability for the HR round
Tips
- Make your resume the core of your preparation: every project and every method listed will be questioned in depth.
- Review classic ML thoroughly, especially tree ensembles, logistic regression, SVM and regularization; these come up more than newer topics.
- Keep medium-level algorithm practice going, with extra dynamic programming and backtracking.
- Find out which business group the role sits in and read about its products; questions and cases follow the group's business.
- If you can, get an employee referral; some referred candidates skipped the written test.
- For LLM and NLP algorithm roles, add transformer internals and the basics of RL-based tuning.
- Expect two to four technical rounds, possibly spread over weeks, and follow up politely if feedback is slow.
- Be on time and prepared for a final HR discussion of pay and start date.
See the rest of the Tencent loop
The later stages, the questions reported for them and how to prepare are in Pro.
Unlock with ProPractice the Tencent loop
Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.
Tencent interview FAQ
How many interview rounds does Tencent have for ML roles?
Candidates report 7 stages: Online written test, Resume and project deep dive, ML and deep learning fundamentals, Live coding, Business and ML case (data science and applied roles), Group lead or hiring manager round and HR interview. Application or employee referral, then for many roles an online written test of algorithm problems, sometimes with multiple-choice math, logic or situational questions. Next come two to four technical interviews by phone or video, each about an hour, which open with a self-introduction and move between your resume projects, ML fundamentals and a live coding problem; data science rounds often add a short business case tied to the group's product. One of the later rounds is usually with the group lead, a senior engineer or the hiring manager, and an HR interview on motivation, pay and availability closes the loop. Referred candidates sometimes skip the written test. Round counts vary by group, from two interviews to six technical rounds plus a manager round in one US loop. Reported timelines run from about two weeks to two months. This outline rests on 25 first-hand reports, most from before 2024, so ask your recruiter which steps your group uses.
How long does the Tencent ML interview process take?
Reported timelines run About 2 weeks to 2 months.
What does Tencent ask in ML interviews?
By reported question mix: Resume & project deep dive 30%; ML & DL fundamentals 27%; Coding (written test and live) 25%; Data & business cases 9%; HR & behavioral 9%. The most repeated themes: Walk through a resume project and defend every modeling choice; How XGBoost works and how it compares with LightGBM or random forest; Logistic regression: why use it, and how it differs from SVM.
How should I prepare for the Tencent ML interview?
Make your resume the core of your preparation: every project and every method listed will be questioned in depth. Review classic ML thoroughly, especially tree ensembles, logistic regression, SVM and regularization; these come up more than newer topics. Keep medium-level algorithm practice going, with extra dynamic programming and backtracking.
Are there free Tencent interview questions?
Yes. LeetML has 2 questions reported from Tencent interviews, and 1 of them are free to solve in the browser; they are listed by stage above.
Related interview guides
- ByteDanceSequential one-hour rounds that pair a deep resume grilling with a classic algorithm problem and ML fundamentals.
- AlibabaResume-driven technical rounds that mix ML fundamentals, hand-written code and a hard look at your own projects, then HR.
- BaiduResume-led technical rounds that mix ML fundamentals, a little math and classic data-structure coding.
- HuaweiResearch deep dives, ML fundamentals and short algorithm coding, with a loop that changes a lot by region and lab.
- DeepSeekResearch- and systems-heavy loop: know the company's own papers, build model parts from scratch and reason about training at scale.
- Z.ai (Zhipu AI)Deep questions on GLM-style architecture and post-training, RL and GUI-agent design, and PyTorch implementation; very few reports so far.