Skip to content
中文Sign in

Every question they ask,
AI helps you think it through

In a mock interview, seconds after each question you see what it tests, the steps to answer in and what to avoid. Prep before, review after, until you can handle the real one yourself

No sign-in · Paste a JD and get this round's prep in half a minute

Example: a mock interview in progress. The interviewer asks how a 12-point drop in handoff rate was calculated. The live hint card beside it is already written: the question is a metric check, the interviewer is confirming the number holds up, the answer chain is to define the metric first, then cover the scope and how noise was ruled out side by side, then give the number, with a phrase to avoid. The interviewer then asks how the eval set was built and which of two metrics to protect, and a new hint card appears for each question

Example · In the AI mock interview, every question can get a hint card

One round, four steps

Scroll to see what each step does and what you get. No sign-in before your first result

01 · Half a minute

Paste the JD, attach your résumé

Paste the whole posting and drop in your résumé. No interview type to pick, no form to fill

ResultBefore the AI says a word, the JD is split into requirements, with the ones most likely to be pushed on marked
02 · About half a minute

Get your prep for this round

Written section by section from your résumé and real coaching notes, and you watch it being written

ResultWhat this round tests, which questions come up, how to answer each, where points get lost
03 · About a quarter hour · optional

Run a mock if you want to

AI plays this round's interviewer, asks from the prep and always follows up on vague answers. Type or talk; skip it if you'd rather go straight to the interview

ResultAn interview that feels like the real round. Stuck? Turn on hint cards and see what the question tests and the steps to answer in within seconds
04 · Right after

Read the report, fix the line that lost points

Every lost point is struck out and rewritten, plus the three things to do before the interview

ResultYou know which sentence to change and what to say, then redo only the questions you missed
Role: AI Product Manager (shopping assistant)Duties: own the assistant roadmap, from pre-sale questions to after-sale supportDefine the assistant's quality metrics and evaluation: eval sets, bad cases, launch barWork with ML and ops to balance quality, latency and model costRequirements: shipped an LLM product from 0 to 1, can explain how numbers are defined
Lin_Ke_resume.pdfRound 1 ▾See what's tested →

7 requirements · the 3 in green get pushed on most

LLM from 0 to 1EvaluationDefining numbersQuality vs costCross-team workShopping3+ years

ResultBefore any prep is written, reading the JD already shows what's tested

Interview prep · Round 1

This round, they want proof you've shipped an LLM product

  1. 1Did you build it, or watch from the side?They'll ask what you did in the last week before launch
  2. 2Did you build the evaluation system?Where the eval set came from, who labeled it, where the bar is
  3. 3Can you define your numbers in one breath?"40% resolution lift" will be taken apart: denominator, baseline, relative or absolute
Know the businessConversion funnelCost of a wrong answerCost per call
Took 26s
  1. Read JD7 requirements
  2. Match résumé3 experiences
  3. Coaching notes14 judgments
  4. Write prep6 sections

ResultPrep written for your résumé: what's tested, the business to know, likely questions, where points get lost

Round 1 mock · interviewer: peer PM3/5
I

Interviewer

Tell me about an LLM product you took from 0 to 1

Y

You

I led the support bot launch, started with the five biggest question types, built the eval set myself, and the results were pretty good

I

Interviewer Follow-up

You said "pretty good". What's the number, and what's the denominator?

Listening… or type

ResultLike this round's real interviewer: questions from your résumé, follow-ups on anything vague

3.2/ 5
Structure3
Credibility4
Depth3
Numbers2
Delivery4
Q2 · What you said → what to say

I led the support bot launch and started with the five biggest question types, and the results were pretty good resolution went from 31% to 43%, out of all conversations in 24h

Lost points"Pretty good" has no number, so they have to push

ResultEvery lost point struck out and rewritten; next, redo only what you missed

Paste the JD, attach your résumé

Paste the whole posting and drop in your résumé. No interview type to pick, no form to fill

Same question, here's the difference

Generic AI on the left, OfferSpar's judgment on the right. Drag the line in the middle

Round 2 · the business owner asks

"Your résumé says resolution went up 40%. How did you calculate that?"

Example
Generic AI

When answering this kind of question, use the STAR method: describe the project background, explain your task and the actions you took, then quantify the result with data. Highlight your analytical skills and results orientation, keep it clear and logical, stay confident, and show deep business understanding and teamwork

OfferSpar

They don't want a prettier number. They're checking whether you actually did the work. Next they'll ask three things in a row: what's the denominator, what's the baseline, and is 40% relative or absolute

So say the definition in your first sentence: share of conversations with no human handoff and no repeat contact within 24 hours, from 31% to 43%, 40% relative and 12 points absolute, over a three-week rollout of about 18,000 conversations. After that, every follow-up is easy

← Generic AIOfferSpar →

Where the judgments come from

The coaching team has worked one-on-one with hundreds of candidates. What interviewers are really checking, session after session, has been written down, and OfferSpar draws on it before it tells you anything

619+students landed offers with the team's coaching
39%average pay raise across those offers
42offers at ¥1M+ a year

Results from the Qingheng team's coached students, as of 2026-10. These are the team's coaching results, not usage stats for this tool

Try the whole thing free, then pay as you go

Sign up for one full free run: prep, a mock interview, the report and a review. After that, buy credits, charged by actual usage: you see roughly how many before you start. No subscription, no auto-renewal

Top-up

¥19.9

36 credits · Valid for 30 days

About 3 full runs of prep → mock → debrief, or about 1 hour of voice practice

Standard

¥39

72 credits · Valid for 90 days

About 6 full runs of prep → mock → debrief, or about 2 hours of voice practice

Sprint Best value

¥99

180 credits · Valid for 180 days

About 16 full runs of prep → mock → debrief, or about 5 hours of voice practice

Long run

¥299

550 credits · Valid for 365 days

About 50 full runs of prep → mock → debrief, or about 15 hours of voice practice

Full pricing and what things usually cost

Read one first, or check yourself

Free interview articles and tools. No sign-in

Free tools

Help and updates

  • Help centerOne page per feature: where to find it, how to use it and what to watch out for
  • ChangelogWhat we shipped, described by what you will notice, newest first

Questions

Anything else? Ask us through Feedback at the bottom right of any page

Is my resume and interview data safe?

Data is stored on servers in mainland China and only you can see it. Mock interview audio is never stored, interview recordings you upload are deleted by default once transcribed, your dossiers are never shown to other users or published, and you can delete your resumes, jobs, records and account memory at any time. Details in the Privacy Policy.

Will the AI make up experience for me?

No. Every suggestion is based only on the experience and numbers actually on your resume. Where nothing fits, it tells you how to talk about it honestly. We won't change your degree or title or invent projects either; a background check would catch it.

Can I turn the AI interviewer off?

Yes. Interview practice has a switch at the top, AI interviewer. On, the AI plays the interviewer, asks and follows up, and every question can get a hint card. Off, it listens to the interviewer, transcribes in real time, gives a hint card for each new question and a debrief at the end. It works for any round, any time.

What does "judgment from real coaching" mean?

Our coaches turn the calls they make again and again into coaching cards: what a question is really checking, which answers get grilled, which claims sound made up. No raw material enters the product. Cards are stripped of anything that could identify a person and reviewed before going live. The AI draws on them when writing prep and debriefs.

How is this different from a human coach? Does it replace one?

Not entirely. The tool is for frequent, everyday prep and debriefs: every round, any time. A human coach goes deeper on positioning, resume rewrites and face-to-face mock feedback. the Qingheng team offers human coaching too, and Qingheng cohort course students use this tool for free.

Which roles does it support?

It's strongest for AI product managers, product managers, Forward Deployed Engineers and new grads. Other roles work too, since general judgment on self-introductions, project deep-dives, HR rounds and pay talks applies everywhere, but role-specific judgment is still growing.

Does it work for interviews in English?

The interface is in English and Chinese. Right now the AI writes prep and debriefs mainly in Chinese, and judgment for English-language interviews is still being built, so don't rely on it alone for English interviews yet.

What is account memory? Can I delete it?

After each debrief, the AI keeps what's useful long-term, such as strengths, weak spots and points you often get asked about, and refers to it next time. You can view, edit or delete it any time.

Can I use it now? Does it cost anything?

Yes. A new account comes with one free trial: go through prep, a mock interview and a debrief once. To keep going, buy a credit pack and pay only for what you use. Qingheng cohort course students get credits from the Qingheng team.

Is the AI always right?

No. It's practice material and can be wrong, so weigh its debriefs against your own situation. AI-generated content is labeled in the product.

When's your next interview?

Paste the JD now and you'll still have time to read the prep; run a mock too if you want

Paste a JD to start