AI Interview Prep Checklist (2 Weeks Out)

Week 1: Fundamentals + practice

Day 1-2: ML theory refresh

Cover:

• Bias-variance tradeoff (be able to diagnose from learning curves)
• Regularization (L1 vs L2, dropout, weight decay in AdamW)
• Overfitting prevention (data, regularization, early stopping, capacity reduction)
• Evaluation metrics (PR-AUC vs ROC-AUC, when each fails)
• Gradient descent variants (SGD vs Adam vs AdamW)

Use the AIKnowHub ML Theory questions at /interviews to test yourself.

Day 3-4: LLMs + applied AI

Cover:

• Transformer architecture (attention, multi-head, positional encoding)
• LLM API basics (context windows, streaming, function calling)
• RAG (chunking, embedding, retrieval, reranking, citation)
• Fine-tuning approaches (full, LoRA, RLHF, DPO)
• Agents (tool design, safety, loops)
• Evals (eval set design, LLM-as-judge)

Apply: walk through "Build an AI Docs Chatbot" workflow mentally. Could you explain it end-to-end with diagrams?

Day 5: System design — 5 walkthroughs

Pick 5 system design prompts. For each:

• Spend 60 minutes designing (out loud, in writing or whiteboard)
• Capture: clarifying questions, architecture sketch, model choices, scale numbers, cost, monitoring

Suggested set:

1. RAG chatbot
2. Recommendation engine
3. Fraud detection
4. AI agent (meeting booking)
5. Evals platform

See AIKnowHub /interviews → System Design for example questions + answers.

Day 6: Behavioral stories

Write 5-7 stories in STAR-but-tighter format:

1. Hardest technical problem solved
2. Time you disagreed with team
3. Failure and what you learned
4. Cross-functional collaboration
5. Ethics / safety judgment call
6. Time you mentored or helped someone
7. Why this company / role specifically

Each story: 90-150 seconds spoken. Practice out loud.

Day 7: Coding practice

If targeting FAANG: 5-10 LeetCode mediums in a topic you're rusty on (arrays, hash maps, graphs).

If targeting AI startups: implement from scratch:

• Scaled dot-product attention in NumPy
• A tiny RAG (load docs → embed → retrieve → generate)
• A simple agent loop with one tool

For all: practice in a shared editor (CoderPad / HackerRank), not your IDE.

Week 2: Mocks + gap closing

Day 8: Mock 1 — coding

With: a friend, peer, or paid service (interviewing.io, Pramp). Use a real interview problem.

Debrief:

• Where did you go silent?
• Did you clarify before coding?
• Did you handle edge cases?
• Did you talk through tradeoffs?

Take detailed notes. The mock is data, not just practice.

Day 9: Mock 2 — system design

With: someone senior to you, ideally in an AI role. Pick a prompt you haven't done.

Score yourself on:

• Clarifying questions (5+ is good)
• Time management (didn't run out)
• Tradeoff articulation
• Awareness of failure modes
• Practical realism (not over-engineering)

Day 10: Targeted gap closing

Based on mocks, identify your weakest area. Spend the day on it:

• If coding was weak: more coding practice in that area
• If system design was weak: 2 more walkthroughs + read a senior engineer's blog
• If behavioral was weak: rewrite 2-3 stories that fell flat

Day 11: Mock 3 — behavioral + executive

With: someone in the role you're targeting. Test:

• Tell me about yourself (2 min)
• Why this company
• 3 STAR stories
• Asking thoughtful questions back

Watch for: rambling, vague outcomes, generic answers.

Day 12-13: Targeted review

• Re-read the company's blog / engineering posts
• Review job description; map your projects to needs
• Refine 2-3 stories that match what they need
• Prep questions for THEM (3-5 per interviewer if possible)

Day 14: Rest

Seriously. Light review only. Cramming hurts performance.

• Walk
• Sleep 8 hours
• Eat well
• Re-read your projects so they're top of mind
• Prepare logistics for the next day (laptop charged, water bottle, backup internet, etc.)

The day-of checklist

• 8+ hours sleep
• Eat a real meal
• 5 minutes to skim your notes (not cram)
• Quiet space; backup internet plan
• Water + snack
• Your portfolio + resume open in a tab
• Pen + paper for system design

What NOT to do in the last week

• Don't learn new frameworks
• Don't read new papers you can't discuss deeply
• Don't try to memorize complex algorithms you weren't already comfortable with
• Don't apply to more jobs (focus on this one)
• Don't skip sleep to study more

After the interview

• Write a quick thank-you (24 hours)
• Note what surprised you / went well / went badly
• Move on; don't refresh your inbox

Most candidates lose offers in the last week through cramming, lack of sleep, or skipping mocks. If you've done the work in week 1, week 2 is consolidation, not new acquisition.