Mock Interview Playbook
A four-week plan, an AI mock-interviewer prompt, a scoring rubric, three scenarios and a day-of checklist.
Practice makes interviews easier. Follow the four-week plan, run mock interviews with a friend or an AI interviewer, score yourself on the same checklist each time and fix your weakest answers first. On the day, ask questions, think aloud and be honest about what you do not know.
- Mock interview
- A practice interview that feels like the real one.
- Rubric
- A checklist used to score your answer.
- Debrief
- Reviewing what went well and badly afterwards.
Follow the four-week plan. On the day, ask questions and think aloud.
Knowing the material is not the same as performing under interview conditions. This last lesson is a practice plan. It has four weeks of steady preparation. It has mock-interview scenarios that you can run with a friend or an AI assistant. It has a scoring rubric, so that you improve on purpose. And it has a checklist for the day itself.
A four-week plan
Plan short sessions of 45 to 90 minutes. Put them in your calendar. Regular practice beats long, rare cramming.
- Week 1, fundamentals and prompting: study lessons 1 and 2 of this path. Record your answers and redo them until they are well structured and under two minutes. Prepare your five stories.
- Week 2, RAG, agents and protocols: study lessons 3 to 5. Draw one RAG design and one agent design against a timer. Do coding drills 1 to 3.
- Week 3, security, evaluation, production and cloud: study lessons 6 to 8. Do the worked system-design answer cold, then a different design. Do coding drills 4 to 6.
- Week 4, staff topics and mocks: study lesson 9. Do at least three full mock interviews with feedback. Polish your portfolio case studies. Prepare questions for the interviewer.
Run a mock interview with an AI assistant
An AI assistant can play a tough interviewer. Give it the prompt below. Tell it to ask one question at a time and to give feedback only after you answer. Do not let it write your answers for you.
Act as a senior interviewer for a {{role}} role at a {{company_type}} company. Run a {{minutes}}-minute system-design interview.
Rules:
- Ask one question at a time and wait for my answer. Start with: "{{design_question}}".
- Act like a real interviewer: ask clarifying follow-ups, push back on weak assumptions, and probe failure modes, security, evaluation, cost and scale when I skip them.
- Do not give me the answer. If I'm stuck for more than a minute, give a small hint.
- At the end, score me 1-5 on: requirements and clarification, architecture quality, failure modes, security, evaluation and observability, trade-offs and communication.
- Give the three most important things to improve, and one thing I did well. Be direct, not flattering.Scoring rubric for your own practice
After every mock, score yourself on each line. Be honest. A score below four tells you what to practice next.
Score each 1-5 after every mock. Aim for 4+ in each before a real interview. 1 Clarification & requirements asked about users, scale, cost of error, data sensitivity, latency; stated NFRs 2 Architecture & baseline simplest viable design first; added components only for stated requirements 3 Failure modes named at least 5 with detection and recovery 4 Security trust boundaries, injection, authorization outside the model, tenant isolation 5 Evaluation & observability metrics, thresholds, traces, release gates 6 Trade-offs & communication alternatives rejected with reasons; structured, calm, honest about uncertainty
Three mock scenarios
Run each one for 40 minutes. Practice the same structure every time.
- An internal HR-policy assistant for 20,000 employees in three countries. It holds confidential documents, and each role has different access rights.
- An autonomous agent that sorts and answers customer emails for an online shop. It can issue small refunds.
- A coding agent that opens pull requests against a company's private code.
After each mock
Write down every question you were asked. Mark your weakest answer. Write the top three fixes. Redo the weakest answer the next day, aloud. Compare the two recordings. This small habit is where most of the improvement comes from. A mock without a debrief is only a rehearsal.
Day-of checklist
- Reread your five stories and the answer structure. Do not cram new topics at the last minute.
- Test your audio, video, drawing tool and internet. Keep water, a pen and paper nearby.
- Clarify before you design. Think aloud. State your assumptions. Draw the data flow.
- If you do not know something, say what you would check and how. Honesty with a method beats bluffing.
- Ask your own questions: how success is measured, how they test and monitor AI features, who owns incidents, and what the team's biggest AI risk is.
- Afterwards, write down every question and your weakest answer within an hour.
- ✓Deliberate practice beats rereading. Use timed answers, recorded mocks and rubric scores.
- ✓Use an AI assistant as a tough interviewer with a fixed rubric. It must not give you the answers.
- ✓Run three different system-design scenarios until your structure feels automatic.
- ✓Have questions ready for the interviewer, and debrief within the hour.
Schedule your four weeks
- Put the four-week plan in your calendar with specific sessions of 45 to 90 minutes and one mock each week.
- Run mock 1 with the AI interviewer prompt on scenario 1. Record your rubric scores.
- Ask a friend or mentor to run mock 2 live. Compare their scores with yours.
- After each mock, write the top three fixes. Redo the weakest answer the next day.
Q1Why score yourself with a rubric after each mock?show answer
In plain words: It shows your weak spots in numbers, so you practise the right thing and see your progress.
Full answer · It turns vague impressions into specific weaknesses you can drill (for example, you always skip failure modes), and it shows progress over time.
Q2What should you do when you don't know an answer in an interview?show answer
In plain words: Say so honestly, explain how you would find out and give your best reasoned guess. A clear method beats bluffing.
Full answer · Say so plainly, state what you'd check and how you'd reason it out, and give your best-reasoned view with caveats. A method and honesty beat a confident bluff.
Q3What should you ask the interviewer at the end?show answer
In plain words: Ask how they test and watch their AI. Ask who handles problems and how they judge risk. Ask what their biggest challenge is.
Full answer · Questions that reveal reality: how they evaluate and monitor AI features, how incidents are handled and who owns them, how they decide autonomy and risk, and what the team's biggest AI challenge is right now.