Interview Loop Strategist

Interview Loop Strategist

End-to-end orchestrator for senior-level AI/ML interview preparation. Coordinates timelines, story coherence across rounds, mock interview cadence, energy management, and post-interview debrief -- routing each round type to its specialist skill.

When to Use

Use for:

• Building a complete interview prep plan for a specific company
• Generating 2-week, 1-month, or 2-month preparation timelines
• Ensuring story coherence -- the same project told correctly across behavioral, technical, and HM rounds
• Scheduling and tracking mock interview cadence
• Energy management strategy for all-day virtual or onsite loops
• Post-interview debrief analysis and improvement planning
• Coordinating across all 7 round-specific interview skills

NOT for:

• Resume or CV creation (use cv-creator)
• Career narrative extraction (use career-biographer)
• Practicing a single round type in isolation (use the round-specific skill directly)
• Salary negotiation or offer evaluation
• General career counseling

---

Full Interview Pipeline

flowchart TD
    A[Career Biographer] -->|Extracts narrative| B[CV Creator]
    B -->|Resume finalized| C[Interview Loop Strategist]
    C --> D{Company Target Selected}
    D --> E[Generate Prep Timeline]
    E --> F[Story Coherence Matrix]
    F --> G[Round-Specific Prep]

    G --> G1[Recruiter Screen<br/>Self-prep]
    G --> G2[CodeSignal / Coding<br/>senior-coding-interview]
    G --> G3[Hiring Manager Screen<br/>hiring-manager-deep-dive]
    G --> G4[ML System Design<br/>ml-system-design-interview]
    G --> G5[Technical Deep Dive<br/>anthropic-technical-deep-dive]
    G --> G6[Tech Presentation<br/>tech-presentation-interview]
    G --> G7[Values & Behavioral<br/>values-behavioral-interview]

    G1 --> H[Mock Interviews<br/>interview-simulator]
    G2 --> H
    G3 --> H
    G4 --> H
    G5 --> H
    G6 --> H
    G7 --> H

    H --> I[Debrief & Adjust]
    I -->|Iterate| G
    I --> J[Interview Day<br/>Energy Protocol]
    J --> K[Post-Loop Debrief]
    K --> L{Offer?}
    L -->|Yes| M[Negotiation Phase]
    L -->|No| N[Gap Analysis & Retry]
    N -->|Update plan| E

---

Skill Routing Table

Each round type maps to a specialist skill. The strategist coordinates -- it does not execute round-specific practice.

---

Prep Timeline Selection

Choose based on time until first round:

flowchart LR
    T{Time Available?}
    T -->|< 2 weeks| P1[Intensive Plan<br/&gt;4-6 hrs/day]
    T -->|2-5 weeks| P2[Balanced Plan<br/&gt;2-3 hrs/day]
    T -->|6+ weeks| P3[Thorough Plan<br/&gt;1-2 hrs/day]
    P1 --> R[See references/<br/>preparation-timeline-templates.md]
    P2 --> R
    P3 --> R

For detailed daily schedules, consult references/preparation-timeline-templates.md.

---

Story Coherence Matrix

A senior candidate has 5-8 strong projects. Each project will surface in multiple rounds but must be tailored to the audience and evaluation criteria of that round.

How to Build the Matrix

1. List top 5 projects from career-biographer output (or direct input) 2. For each project, write 3 versions:

3. Cross-check for contradictions -- dates, team sizes, your role, metrics must be identical across versions 4. Map projects to Anthropic values -- which project demonstrates which value (safety, honesty, broad benefit)?

Example Coherence Entry

Project: Real-Time Object Detection Pipeline (2019-2022)

---

Energy Management Protocol

All-day interview loops (4-6 hours) are endurance events. Cognitive fatigue causes more failures than knowledge gaps.

Before Interview Day

• Sleep 7-8 hours for 3 consecutive nights prior (not just the night before)
• Prepare environment: quiet room, backup internet, charged devices, water, snacks
• Do NOT cram the morning of -- review only your story coherence matrix and 1-page cheat sheet
• Light exercise the morning of (walk, stretch -- not a hard workout)

During the Loop

Cognitive Sequencing

If you can influence round order (sometimes companies ask preference):

1. Start with your strongest round -- builds confidence momentum 2. Put coding early -- requires peak cognitive freshness 3. Behavioral/values in the middle -- these are conversational and let you recover 4. Presentation whenever you rehearsed it -- muscle memory carries you 5. Avoid technical deep dive as last round -- fatigue makes it easy to ramble

---

Debrief Framework

Run after every mock AND every real interview round.

Immediate (within 30 minutes)

1. Dump raw notes -- what questions were asked, what you said, what you wish you said 2. Emotional check -- how did you feel? Confident, uncertain, surprised? 3. Time check -- did you run over? Under? Where did you lose time?

Structured Analysis (within 24 hours)

Pattern Detection (weekly)

• What round types consistently score lowest?
• Which stories land well? Which fall flat?
• Are you improving on last week's action items?
• Adjust prep timeline allocation based on weakness trends

For scoring rubrics per round type, consult references/mock-interview-rubrics.md.

---

Anti-Patterns

Anti-Pattern: Uniform Preparation

Novice: Spends equal time on every round type -- 2 hours coding, 2 hours design, 2 hours behavioral, repeat. Expert: Analyzes personal weaknesses and round weighting. A candidate who aces design but freezes in coding allocates 60% of prep to coding. A candidate whose stories are inconsistent spends dedicated time on the coherence matrix. Detection: Prep log shows identical hours across all categories despite known weaknesses.

Anti-Pattern: Isolation Prep

Novice: Prepares each round independently. Tells a behavioral story about leading a team of 6 in one round, then says "I was the sole contributor" for the same project in a technical round. Expert: Uses the story coherence matrix to thread a consistent narrative across all rounds. Reviews the matrix before every mock. Has a peer check for contradictions. Detection: Same project described with conflicting details (team size, timeline, your role, metrics) across different round types.

Anti-Pattern: Mock Avoidance

Novice: Reads interview guides, watches YouTube videos, reviews flashcards -- but never actually practices speaking answers aloud under time pressure. Expert: Runs at minimum 2 full mock interviews per week in the final month. Records mocks. Reviews recordings. Uses interview-simulator for structured scoring. Treats mocks as the primary prep activity, not supplementary. Detection: Zero mock session recordings in history. Unable to answer questions within time limits despite "knowing the material."

---

Anthropic-Specific Notes

Anthropic's interview process (as of early 2026) emphasizes:

1. AI Safety understanding -- not just technical competence, but genuine engagement with alignment, interpretability, and responsible deployment 2. Technical depth over breadth -- they want to see how deep you can go on your own work, not surface-level familiarity with everything 3. Collaborative problem-solving -- interviews are designed to feel like working sessions, not interrogations 4. Intellectual honesty -- saying "I don't know" or "I was wrong about that" is valued over bluffing 5. Mission alignment -- why Anthropic specifically, not just "any AI company"

For detailed company-specific loop structures (Anthropic, Google DeepMind, OpenAI, Meta FAIR), consult references/company-specific-loops.md.

---

Process: First Session with a Candidate

1. Assess current state: Which company? When is the interview? What's your background? 2. Review upstream artifacts: career-biographer output, cv-creator resume, any existing prep 3. Select timeline template: 2-week / 1-month / 2-month based on available time 4. Build story coherence matrix: Top 5 projects x 3 versions each 5. Identify weakness areas: Self-assessment + any prior interview feedback 6. Generate personalized prep plan: Daily schedule with skill routing 7. Schedule first mock: Within 48 hours of starting prep 8. Set debrief cadence: After every mock, weekly pattern review

---

Reference Files

Company-Specific Interview Loops

Detailed interview loop structures for top AI companies, tailored for senior ML/AI/CV/NLP engineers (Staff/Principal level, 10-15+ years experience). Information reflects 2025-2026 practices.

Caveat: Interview processes change. Always verify current structure with your recruiter. This document provides a strong baseline for preparation focus.

---

Anthropic

Loop Structure (2026)

Anthropic's process for senior technical roles typically follows this sequence:

Application/Referral
    |
    v
Recruiter Screen (30 min, phone)
    |
    v
CodeSignal Assessment (70 min, async)
    |
    v
Hiring Manager Screen (45 min, video)
    |
    v
Full Loop - Virtual Onsite (4-5 hours, single day)
  ├── Coding Interview (60 min)
  ├── ML System Design (60 min)
  ├── Technical Deep Dive (60 min)
  ├── Tech Presentation (45-60 min)
  └── Values & Culture Fit (45 min)
    |
    v
Team Match Conversations (1-2 calls, 30 min each)
    |
    v
Offer / Debrief

Typical timeline: 3-6 weeks from application to offer.

Recruiter Screen

Duration: 30 minutes Format: Phone or video call with talent team member What they evaluate:

• Motivation for Anthropic specifically (not just "AI is cool")
• High-level technical background confirmation
• Role fit / level calibration
• Logistics: visa, location, timeline, compensation expectations

Preparation priorities:

• Know Anthropic's mission (safety-first AI development) cold. Reference specific things: Constitutional AI, Claude's character training, interpretability research, Responsible Scaling Policy.
• Have a crisp 2-minute pitch: who you are, what you've built, why Anthropic now.
• Know your compensation expectations. Research levels (IC3-IC5 for engineers) and corresponding bands.
• Have thoughtful questions about the team and role.

Common pitfall: Saying "I'm excited about AI" without demonstrating specific knowledge of Anthropic's approach. Every AI company candidate says this. What makes Anthropic different to you?

CodeSignal Assessment

Duration: 70 minutes Format: Async, completed on your schedule within a 7-day window Content: 2-4 algorithmic coding problems, general programming assessment Language: Choice of Python, Java, JavaScript, C++, or similar

What they evaluate:

• Algorithmic problem-solving ability
• Code quality and correctness
• Time management under pressure

Preparation priorities:

• Practice LeetCode medium-to-hard problems (not easy -- the bar is senior)
• Focus on: trees/graphs, dynamic programming, string manipulation, system design components
• Practice in CodeSignal's UI (it's different from your IDE -- test it)
• Time yourself: average 15-20 min per problem

Common pitfall: Treating this as a formality. Senior candidates sometimes underperform on algorithmic coding because they haven't done it recently. Dedicate real prep time here.

Hiring Manager Screen

Duration: 45 minutes Format: Video call with the hiring manager (your potential direct manager) What they evaluate:

• Technical judgment and decision-making
• Leadership style and team collaboration evidence
• Culture add (not culture fit -- they want diversity of thought)
• Motivation for this specific team and role

Preparation priorities:

• Research the hiring manager (publications, talks, team's recent work)
• Prepare 3-4 stories demonstrating: technical leadership, navigating ambiguity, mentoring, disagreeing and committing
• Have a clear answer for "Why this team?" that references their specific projects
• Prepare 3+ thoughtful questions about team dynamics, technical challenges, and growth

Common pitfall: Treating this like a behavioral interview. The HM is evaluating you as a potential team member and direct report. They want to see how you think, not just hear polished stories.

Full Loop: Coding Interview

Duration: 60 minutes Format: Live coding with a senior engineer, shared editor or whiteboard Content: 1-2 algorithmic/systems problems, harder than CodeSignal

What they evaluate:

• Problem decomposition and approach articulation
• Code quality under pressure
• Communication during problem-solving
• Ability to handle hints and pivot when stuck

Anthropic-specific notes:

• Interviewers are collaborative. They will hint and guide -- this is intentional. Take hints gracefully.
• They care as much about how you communicate your approach as the solution itself.
• Python is the dominant language internally. Using Python is slightly advantageous.

Full Loop: ML System Design

Duration: 60 minutes Format: Whiteboard/video discussion with a senior ML engineer Content: Design an end-to-end ML system for a realistic problem

What they evaluate:

• End-to-end systems thinking (data -> training -> serving -> monitoring)
• Awareness of ML-specific production challenges (drift, feedback loops, evaluation)
• Tradeoff analysis and decision-making under ambiguity
• Practical experience with real ML systems at scale

Anthropic-specific notes:

• Problems may relate to language model training, RLHF, evaluation, or safety
• Deep understanding of LLM architectures, training dynamics, and evaluation is expected at the senior level
• They value practical experience over theoretical knowledge
• Discussing failure modes and safety considerations is a strong signal

Full Loop: Technical Deep Dive

Duration: 60 minutes Format: Discussion-based with senior technical staff Content: Deep forensic examination of 1-2 of your past projects

What they evaluate:

• True depth of technical understanding
• Intellectual honesty (what you don't know, what went wrong)
• Ability to explain complex systems clearly
• Quality of technical judgment in past decisions

Anthropic-specific notes:

• This is often the most heavily weighted round for senior candidates
• They will go extremely deep. Expect questions about decisions you made 3-5 levels deep: "Why that architecture? What alternatives did you consider? What would you do differently? What were the failure modes?"
• Prepare to discuss one project for the full 60 minutes. They will exhaust your knowledge.
• Intellectual honesty is paramount. Bluffing is the fastest way to a rejection.

Full Loop: Tech Presentation

Duration: 45-60 minutes (30 min presentation + 15-30 min Q&A) Format: You present a technical topic to 3-5 engineers Content: A talk about your most significant technical work

What they evaluate:

• Communication ability (can you explain complex ideas clearly?)
• Technical substance (is this real work with real depth?)
• Audience calibration (appropriate level for senior ML engineers?)
• Q&A handling (how do you respond to challenges and curiosity?)

Anthropic-specific notes:

• Pick work that genuinely excites you. Enthusiasm is visible and valued.
• Audience is extremely technical. Don't oversimplify -- they can handle it.
• Expect hard questions. The Q&A is evaluative, not just polite interest.
• Connecting your work to AI safety or responsible development (where genuine) is a positive signal.

Full Loop: Values & Culture Fit

Duration: 45 minutes Format: Conversation with a cross-functional team member (possibly non-engineering) Content: Behavioral questions focused on values alignment

What they evaluate:

• Alignment with Anthropic's values: safety, honesty, humility, collaboration
• How you handle ethical dilemmas and ambiguity
• Genuine interest in the mission (beyond compensation)
• How you treat people, especially under pressure

Anthropic-specific notes:

• This is not a checkbox exercise. Anthropic takes values fit very seriously.
• Be prepared to discuss: AI safety concerns you have, times you prioritized safety over speed, how you handle disagreements about technical direction
• They want to see genuine engagement with hard questions, not rehearsed answers
• Reference the Responsible Scaling Policy and Claude's character if relevant (but only if you've actually read them)

Team Match

Duration: 1-2 calls, 30 minutes each Format: Informal conversations with potential team members Purpose: Mutual evaluation of team fit (not scored in the traditional sense)

What to do:

• Be yourself. This is genuinely bidirectional.
• Ask about day-to-day work, team dynamics, biggest challenges
• Evaluate if this is a team where you'd thrive
• It's okay to express preferences if you're matched with multiple teams

---

Google DeepMind

Loop Structure (2025-2026)

Application/Referral
    |
    v
Recruiter Screen (30 min)
    |
    v
Phone Screen: Coding (45 min) OR Research Discussion (45 min)
    |
    v
Full Loop - Virtual or Onsite (5-6 hours)
  ├── Coding Interview #1 (45 min)
  ├── Coding Interview #2 (45 min)
  ├── ML/Research Design (60 min)
  ├── Technical Deep Dive / Past Work (45 min)
  └── Googliness & Leadership (45 min)
    |
    v
Hiring Committee Review
    |
    v
Team Match
    |
    v
Senior Leadership Review (L6+)
    |
    v
Offer

Typical timeline: 6-12 weeks (notoriously slow -- hiring committee adds weeks).

Key Differences from Anthropic

Preparation Adjustments

• Double down on coding: Two dedicated coding rounds means algorithmic weaknesses are more likely to surface. Spend 40%+ of prep on coding.
• Research fluency: If targeting research roles, be prepared to discuss recent papers, propose new research directions, and critique methodologies.
• Googliness: Emphasize collaboration over individual heroism. "How did you help others succeed?" is a common theme.
• Publication record: Not required for applied roles, but a strong publication record significantly strengthens research track applications.

---

OpenAI

Loop Structure (2025-2026)

Application/Referral
    |
    v
Recruiter Screen (30 min)
    |
    v
Technical Phone Screen (60 min) -- coding + ML discussion
    |
    v
Full Loop - Typically Virtual (4-5 hours)
  ├── Coding (60 min)
  ├── System Design / ML Design (60 min)
  ├── Technical Depth / Past Work (60 min)
  └── Manager / Culture Fit (45 min)
    |
    v
Debrief & Decision
    |
    v
Offer

Typical timeline: 3-5 weeks (faster than DeepMind, similar to Anthropic).

Key Differences from Anthropic

Track Differences

Applied Research / Engineering:

• Heavier coding emphasis
• System design focuses on production systems (inference, APIs, scale)
• Past work discussion emphasizes shipping products
• ML knowledge tested at applied level (how to use, not how to derive)

Research:

• Lighter coding (but still tested)
• Research design: propose experiments, discuss methodology, critique papers
• Past work: deep dive into publications and research contributions
• ML knowledge at theoretical depth (loss landscapes, optimization theory, architecture design)

Preparation Adjustments

• Applied track: Focus on coding (LeetCode hard), production ML systems, and shipping under pressure.
• Research track: Focus on paper discussion ability, experiment design, and research taste (which problems are worth solving?).
• Both tracks: Know GPT architecture details. Be prepared to discuss transformer variants, scaling laws, and RLHF at depth.

---

Meta FAIR

Loop Structure (2025-2026)

Application/Referral
    |
    v
Recruiter Screen (30 min)
    |
    v
Phone Screen: Coding (45 min)
    |
    v
Full Loop - Onsite or Virtual (5-6 hours)
  ├── Coding #1 (45 min)
  ├── Coding #2 (45 min)
  ├── ML System Design (60 min)
  ├── Research Discussion / Past Work (60 min)
  └── Behavioral / Leadership (45 min)
    |
    v
Debrief & Decision
    |
    v
Offer

Typical timeline: 4-8 weeks.

Key Differences from Anthropic

Preparation Adjustments

• Coding bar is high: Two coding rounds. Practice on Codeforces/LeetCode hard. Meta's coding interviews trend toward harder algorithmic problems.
• Open source record: Having meaningful open source contributions (especially in ML frameworks like PyTorch) is a significant advantage. Prepare to discuss contributions.
• System design breadth: FAIR interviews may include pure distributed systems design (not just ML pipelines). Know sharding, replication, consensus, and large-scale training infrastructure.
• Publication discussion: Be prepared to present and defend a paper in conversational format. "Walk me through your best paper. What would you do differently?"
• Move fast culture: Stories should emphasize speed, iteration, and impact. "We shipped in X weeks" resonates more than "We spent 6 months perfecting Y."

---

Cross-Company Comparison

Difficulty by Round Type

What Each Company Values Most

Negotiation Leverage Between Offers

If you receive multiple offers, use this knowledge:

• Anthropic and OpenAI compete most directly. An offer from one provides strong leverage with the other.
• DeepMind offers tend to be lower base (Google pay bands) but strong RSUs. Anthropic/OpenAI often counter with higher base + equity.
• Meta FAIR offers are typically highest total comp. Useful as a ceiling anchor in negotiations.
• All four companies respect the other three. A competing offer from any of them is taken seriously.

---

Preparation Priority Matrix

Given limited prep time, how to allocate across companies:

Mock Interview Rubrics

Scoring rubrics for each interview round type. Use after every mock interview (with interview-simulator or a human practice partner) and after real interviews for structured debrief.

---

Universal Scoring Scale

All dimensions use a 1-5 scale:

Target for Anthropic and peers: Average 4.0+ across all dimensions for hire recommendation. A single 2 in any dimension is usually disqualifying.

---

Coding Round Rubric

Duration: 45-60 minutes (typically 1-2 problems) Evaluator focus: Problem-solving approach, code quality, communication

Coding Self-Evaluation Checklist

After each mock coding session:

• [ ] Did I ask at least 3 clarifying questions before starting?
• [ ] Did I discuss my approach and get interviewer buy-in before coding?
• [ ] Did I analyze time/space complexity without being asked?
• [ ] Did I talk through my thought process continuously?
• [ ] Did I test my solution with at least 2 examples (including an edge case)?
• [ ] Did I finish within the time limit?
• [ ] Could I explain every line of my code if asked?

Common Coding Failure Modes

---

ML System Design Rubric

Duration: 45-60 minutes Evaluator focus: End-to-end thinking, tradeoff awareness, practical experience

ML Design Self-Evaluation Checklist

• [ ] Did I spend the first 5 minutes on requirements and scope?
• [ ] Did I draw a clear system diagram (even in text)?
• [ ] Did I discuss both training and serving pipelines?
• [ ] Did I address data collection, labeling, and quality?
• [ ] Did I discuss model evaluation metrics and how to measure success?
• [ ] Did I mention monitoring, drift detection, and feedback loops?
• [ ] Did I present at least one meaningful tradeoff with alternatives?
• [ ] Did I discuss what could go wrong and how to mitigate it?

Common ML Design Failure Modes

---

Technical Deep Dive Rubric

Duration: 45-60 minutes Evaluator focus: Depth of understanding of your own work, intellectual honesty

Deep Dive Self-Evaluation Checklist

• [ ] Can I explain every major technical decision in my top 3 projects?
• [ ] Do I know what I would do differently with hindsight?
• [ ] Can I clearly separate my contributions from the team's?
• [ ] Am I comfortable saying "I don't know" and pivoting to how I'd learn?
• [ ] Can I handle 5 levels of "why" on any decision?
• [ ] Do I have quantified metrics for impact?

---

Behavioral / Values Round Rubric

Duration: 30-45 minutes Evaluator focus: Alignment with company values, self-awareness, collaboration patterns

Behavioral Self-Evaluation Checklist

• [ ] Did each story stay under 3 minutes in the telling?
• [ ] Did I use STAR format (Situation, Task, Action, Result)?
• [ ] Did I include specific, verifiable details?
• [ ] Did I explain my reasoning, not just my actions?
• [ ] Did my stories align with the target company's stated values?
• [ ] Did I demonstrate self-awareness about my growth areas?
• [ ] Did I handle follow-up questions without contradicting my story?

---

Hiring Manager Screen Rubric

Duration: 30-45 minutes Evaluator focus: Leadership, team fit, technical judgment, motivation

---

Tech Presentation Rubric

Duration: 45 minutes (30 min talk + 15 min Q&A) Evaluator focus: Communication, technical depth, audience calibration

Presentation Self-Evaluation Checklist

• [ ] Did I finish within 2 minutes of the target time?
• [ ] Did I start with a hook that made the audience want to listen?
• [ ] Did I clearly explain why this work matters?
• [ ] Were my slides readable from the back of the room (or on a small video call)?
• [ ] Did I handle Q&A questions directly without getting defensive?
• [ ] Did I end with a clear takeaway?

---

Peer/AI Evaluator Guidelines

When using [interview-simulator] or a human practice partner as evaluator:

Before the Mock

1. Share the relevant rubric with the evaluator 2. Specify which round type you're practicing 3. Ask them to take timestamped notes during the mock 4. Request they score each dimension immediately after (not days later)

During the Mock

• Evaluator should simulate realistic interview pressure (time limits, follow-up questions)
• Evaluator should NOT help or coach during the mock -- only evaluate
• If the candidate is stuck, the evaluator can provide a small hint (as a real interviewer would) but should note that a hint was needed

After the Mock

1. Evaluator shares scores (1-5 per dimension) with brief justification 2. Evaluator identifies top 2 strengths and top 2 improvement areas 3. Candidate and evaluator discuss specific moments (use timestamps) 4. Agree on 2-3 concrete action items for next practice session

Calibration Notes

• A score of 3 means "would pass at Anthropic" -- this is a high bar
• First-time mock scores are typically 2-3. This is normal and expected.
• Focus on trend, not absolute score. Going from 2.5 to 3.5 over 4 weeks is excellent progress.
• If scores plateau, change practice approach (different problem types, different evaluator, different round structure)

---

Aggregate Score Tracking

After each mock, compute:

Round Score = average of all dimensions for that round
Overall Score = weighted average across all rounds

Weights (Anthropic):
  Coding:          0.15
  ML Design:       0.20
  Technical Depth: 0.25
  Behavioral:      0.15
  HM Screen:       0.10
  Presentation:    0.15

Target trajectory:

• Week 1: Overall 2.5 (baseline)
• Week 2: Overall 3.0 (foundations)
• Week 3: Overall 3.5 (integration)
• Week 4: Overall 4.0+ (ready)

If you're not tracking toward 4.0 by interview week, consider requesting a timeline extension or focusing exclusively on your lowest-scoring round.

Preparation Timeline Templates

Detailed daily prep schedules for three scenarios. Each plan assumes a senior ML/AI engineer (10-15+ years experience) targeting Anthropic or similar top AI companies.

Convention: Skills referenced in brackets (e.g., [senior-coding-interview]) should be invoked for that session's practice.

---

2-Week Intensive Plan

When to use: Short notice -- recruiter call happened, onsite is in 14 days. Daily commitment: 4-6 hours/day Risk level: High -- no time for weakness remediation, focus on sharpening existing strengths.

Week 1: Foundation & Assessment

Day 1 (Mon): Inventory & Planning

• Morning (2h): Complete story coherence matrix -- list top 5 projects, write technical/impact/narrative versions
• Afternoon (2h): Self-assessment -- rate yourself 1-5 on each round type. Identify top 2 weaknesses.
• Evening (1h): Review company-specific loop structure from references/company-specific-loops.md

Day 2 (Tue): Coding Fundamentals

• Morning (3h): [senior-coding-interview] -- 3 medium/hard problems. Focus on talking through approach before coding.
• Afternoon (2h): Review solutions. Identify pattern gaps (graphs? DP? system design components?).

Day 3 (Wed): ML System Design

• Morning (3h): [ml-system-design-interview] -- Design 1 end-to-end ML system from scratch (recommendation engine or similar).
• Afternoon (2h): Review design against rubric. Practice drawing architecture diagrams while talking.

Day 4 (Thu): Technical Deep Dive

• Morning (2h): [anthropic-technical-deep-dive] -- Prepare 2 projects for deep forensic questioning. Know every decision, tradeoff, and failure.
• Afternoon (2h): Have someone (or [interview-simulator]) grill you on technical decisions. Practice "I don't know, but here's how I'd figure it out."

Day 5 (Fri): Behavioral & Values

• Morning (2h): [values-behavioral-interview] -- Write 8 STAR stories mapped to Anthropic values. Practice delivering each in under 3 minutes.
• Afternoon (2h): [hiring-manager-deep-dive] -- Practice answering "tell me about a time you disagreed with leadership" and similar leadership questions.

Day 6 (Sat): First Full Mock

• Morning (3h): [interview-simulator] -- Run a complete mock loop (coding + design + behavioral). Time each round.
• Afternoon (2h): Debrief using the framework in SKILL.md. Score yourself. Identify top 3 improvement areas.

Day 7 (Sun): Rest

• Light review only -- reread story coherence matrix. No intense practice.

Week 2: Refinement & Performance

Day 8 (Mon): Weakness Drill

• Full day (4h): Focus entirely on your weakest round type from Saturday's mock. Use the appropriate specialist skill.

Day 9 (Tue): Presentation Prep

• Morning (3h): [tech-presentation-interview] -- Build and rehearse your 45-minute technical talk. Time it.
• Afternoon (1h): Practice Q&A -- anticipate the 10 hardest questions about your talk.

Day 10 (Wed): Second Full Mock

• Morning (3h): [interview-simulator] -- Full mock loop with different problems than Day 6.
• Afternoon (2h): Debrief. Compare scores to Day 6. Are you improving on identified weaknesses?

Day 11 (Thu): Story Polish

• Morning (2h): Refine story coherence matrix based on mock feedback. Which stories landed? Which fell flat?
• Afternoon (2h): Practice transitions -- "That reminds me of another project where..." and handling "Tell me more about X" pivots.

Day 12 (Fri): Light Review

• Morning (2h): Review 1-page cheat sheet (key metrics, project timelines, company-specific values).
• Afternoon: Environment prep -- test equipment, prepare backup plans, lay out clothes.

Day 13 (Sat): Final Mock

• Morning (2h): One focused mock on your weakest round type only. Score it.
• Afternoon: Rest. Walk. Don't study.

Day 14 (Sun): Interview Day

• Morning: Light exercise, review cheat sheet, eat well. Execute energy management protocol.

2-Week Milestones

---

1-Month Balanced Plan

When to use: Standard timeline -- you have 4 weeks from recruiter call to onsite. Daily commitment: 2-3 hours/day Risk level: Moderate -- enough time to address 2-3 weakness areas.

Week 1: Foundation

Focus: Story inventory, self-assessment, company research

Week 2: Skill Building

Focus: Dedicated practice on each round type, starting with weakest

Week 3: Integration & Mocks

Focus: Cross-round coherence, presentation, intensive mocking

Week 4: Performance Polish

Focus: Final mocks, confidence building, logistics

1-Month Milestones

---

2-Month Thorough Plan

When to use: Early-stage pipeline -- you've applied and expect 6-8 weeks before onsite. Or you're proactively preparing before even applying. Daily commitment: 1-2 hours/day Risk level: Low -- ample time for deep skill building and multiple iteration cycles.

Phase 1: Weeks 1-2 (Foundation)

Goal: Complete self-assessment, build story inventory, research companies thoroughly.

Week 1:

• Mon-Wed (1h each): Career biography review. Extract top 8 projects with metrics. Use [career-biographer] if not done.
• Thu-Fri (1.5h each): Company research -- read Anthropic's research papers (Constitutional AI, RLHF papers, interpretability), understand product lineup.
• Sat (2h): Write story coherence matrix for top 5 projects.

Week 2:

• Mon-Tue (1.5h each): Self-assessment -- diagnostic coding test (2 hard problems, timed).
• Wed-Thu (1.5h each): Self-assessment -- practice ML design question, practice behavioral stories aloud.
• Fri (1h): Compile weakness ranking. Build personalized plan adjustments.
• Sat (2h): Deep dive into Anthropic's values and culture. Read employee blogs, glassdoor, team pages.

Phase 2: Weeks 3-4 (Skill Building)

Goal: Systematic practice of each round type with focus on top 2 weaknesses.

Week 3:

• Mon-Tue (1.5h each): Coding -- [senior-coding-interview] -- 2 problems/day, focus on approach articulation.
• Wed-Thu (1.5h each): ML Design -- [ml-system-design-interview] -- 1 full design/day.
• Fri (1.5h): Technical deep dive -- [anthropic-technical-deep-dive] -- forensic prep on project #1.
• Sat (2h): First Mock (coding + behavioral only) -- [interview-simulator].

Week 4:

• Mon-Tue (1.5h each): Behavioral -- [values-behavioral-interview] -- refine 8 STAR stories.
• Wed-Thu (1.5h each): HM prep -- [hiring-manager-deep-dive] -- leadership stories, "why this role" narrative.
• Fri (1.5h): Technical deep dive -- [anthropic-technical-deep-dive] -- forensic prep on project #2.
• Sat (2h): Debrief Week 3 mock. Adjust weakness priorities.

Phase 3: Weeks 5-6 (Integration)

Goal: Cross-round coherence, presentation development, regular mocking.

Week 5:

• Mon-Wed (1.5h each): Presentation development -- [tech-presentation-interview] -- build slides, first rehearsal.
• Thu (1.5h): Story coherence audit -- check all 5 projects x 3 versions for contradictions.
• Fri (1.5h): Weakness drill -- whatever scored lowest in Week 3 mock.
• Sat (3h): Second Full Mock -- complete simulated loop -- [interview-simulator].

Week 6:

• Mon (1h): Debrief second mock. Update improvement tracking.
• Tue-Wed (1.5h each): Presentation refinement -- second and third rehearsals.
• Thu-Fri (1.5h each): Coding maintenance -- 1 hard problem/day to stay sharp.
• Sat (2h): Third Full Mock -- focus on weakest round.

Phase 4: Weeks 7-8 (Performance)

Goal: Peak performance, confidence building, logistics.

Week 7:

• Mon-Tue (1.5h each): Final story polish. Practice transitions and pivot handling.
• Wed (2h): Fourth Full Mock -- complete loop with different evaluator/perspective.
• Thu (1h): Debrief. Confirm all round scores are trending 4+/5.
• Fri (1h): Presentation: final rehearsal. Time it to the minute.
• Sat: Light review only. Rest.

Week 8 (Interview Week):

• Mon-Tue (1h each): Light coding for confidence. Review cheat sheet.
• Wed: Environment prep, backup plans, logistics confirmed.
• Thu: Rest. Light exercise. Early dinner. Early bed.
• Fri: Interview Day (or adjust to actual date). Execute energy protocol.
• Sat: Post-loop debrief using SKILL.md framework.

2-Month Milestones

---

Customization Rules

These templates are starting points. Adjust based on:

1. Self-assessment results: If coding is your weakest area, shift 50%+ of practice time there. 2. Company emphasis: Anthropic weights technical depth and values more than raw coding speed. Google DeepMind weights coding more heavily. 3. Prior interview experience: If you've done a loop at a peer company in the last 6 months, you can compress the foundation phase. 4. Available practice partners: If you have senior engineers willing to mock with you, schedule those sessions as immovable anchors and build around them. 5. Energy and sustainability: If 4 hours/day on the 2-week plan causes burnout, drop to 3 hours and accept slightly less coverage. Burned-out performance is worse than underprepared but fresh.

---

Progress Tracking Template

Use this weekly to track improvement:

## Week [N] Progress

### Mock Scores
| Round Type | Score (1-5) | Trend | Notes |
|-----------|-------------|-------|-------|
| Coding | | | |
| ML Design | | | |
| Deep Dive | | | |
| Behavioral | | | |
| Presentation | | | |
| HM Screen | | | |

### Top 3 Improvement Actions This Week
1.
2.
3.

### Story Coherence Check
- [ ] All projects have consistent dates/metrics across versions
- [ ] No contradictions found in this week's mocks
- [ ] At least 1 new story refined based on mock feedback

### Confidence Level: [1-10]
### Hours Invested This Week: [X]
### Next Week Priority: