05 - Interview Sprint Loops
This module gives you a repeatable prep routine for stronger recall and delivery.
Weekly Sprint Structure
Sprint A - Technical Clarity
- Explain one architecture from memory.
- Whiteboard one failure mode and fix sequence.
- Answer one system-design question in 3 minutes.
Sprint B - Story Quality
- Deliver 2 STAR+T stories.
- Add or refine one metric in each story.
- Tighten tradeoff explanation.
Sprint C - Realism and Pressure
- Run 30-minute mock interview.
- Review weak answers and rewrite.
- Re-attempt same questions after 24 hours.
Mock Interview Question Bank
- How do you debug bad RAG answers?
- How do you design reliable tool-calling agents?
- How do you evaluate prompt/model changes?
- How do you balance latency, cost, and quality?
- How do you productionize an LLM workflow?
- How do you use AI coding tools safely?
Add these baseline finals:
- Tell me about an LLM system you built end to end.
- How do you handle tool-call failures in production?
- How would you design an MCP-based coding pilot?
Feedback Rubric
Score each answer from 1 to 5 on:
- Technical depth
- Clarity and structure
- Metrics and measurable impact
- Tradeoff reasoning
- Production realism
Week 4 Alignment
| Day | Use this page for | Deliverable |
|---|---|---|
| Day 25 | Technical drill structure and answer framework | Drill score sheet |
| Day 26 | Behavioral communication and translation | Stakeholder-friendly answer set |
| Day 27 | Full mock loop and revision cycle | Mock transcript and fixes |
| Day 28 | Final launch plan and warm-up cadence | 2-week interview calendar |
Step-by-Step Mock Loop Runbook
| Step | Action | Output |
|---|---|---|
| 1 | Pick 3 technical questions and 2 behavioral questions | Question set |
| 2 | Answer each in 90 seconds | Baseline delivery |
| 3 | Re-answer the weakest one in 3 minutes | Deep-dive version |
| 4 | Score with the rubric | Gap list |
| 5 | Rewrite one weak answer and retry next day | Improvement loop |
Example Code: 90-Second Drill Runner
import time
questions = [
"How do you debug a bad RAG answer?",
"How do you design reliable tool-calling agents?",
"How do you evaluate a prompt change before release?",
]
for question in questions:
print(f"Question: {question}")
print("Start speaking now")
time.sleep(2)
print("Checkpoint: 30 seconds")
time.sleep(2)
print("Checkpoint: 60 seconds")
time.sleep(2)
print("Stop at 90 seconds and score yourself")
Example Scorecard
candidate: self-review
question: How do you debug bad RAG answers?
scores:
technical_depth: 4
clarity: 3
metrics: 2
tradeoffs: 3
production_realism: 4
next_fix: Add one metric and shorten the opening by one sentence.
Interview Q: What is the fastest way to improve weak interview answers?
Model Answer: Record the answer, score it against a fixed rubric, and rewrite only the weakest part instead of starting over. That creates measurable iteration instead of random practice.
Why this matters: This shows you can improve performance with an engineering-style feedback loop.
Interview Q: Why prepare both 90-second and 3-minute versions?
Model Answer: The short version proves clarity and structure, while the longer version proves depth and tradeoff reasoning. Together they prepare you for both recruiter screens and technical interviews.
Why this matters: Interview success depends on adapting answer depth to the round.
Retention Trick: Mistake Ledger
After each mock:
- Write the top 3 misses.
- Write improved phrasing.
- Rehearse corrected answers after 1 day and 1 week.
Weekly Cadence
- Round 1 (Mon/Tue): 90-second clarity answers
- Round 2 (Thu): 3-minute deep dive answers
- Round 3 (Weekend): 30-minute mixed mock loop
Quick Lab (20 min)
Mock loop lab
- Pick any 3 questions from the bank.
- Answer each in 90 seconds.
- Re-answer one question in 3 minutes.
- Grade yourself with the rubric and update one story.