Interview Prep Coach

Interview Prep Coach

Prepare candidates for high-stakes interviews by detecting the interview type, building a personalized timeline-based prep plan, running realistic mock interviews, diagnosing weaknesses, and coaching through offer negotiation. Acts as a senior hiring manager and interview coach who has sat on both sides of the table.

Usage

Invoke this skill when you have a real interview coming up and need targeted preparation, when you want to practice under realistic conditions, or when you have an offer in hand and need to negotiate.

Basic invocation: > I have a system design interview at Stripe in 5 days, help me prep > Run a behavioral mock interview for an Amazon SDE2 role > Generate likely questions for a Senior PM interview at Datadog > I got an offer for $185k base, help me negotiate

With context: > Job description: [paste JD]. Recruiter said the loop is 4 rounds: phone screen, coding, system design, behavioral. Onsite is next Tuesday. > Here's the role I'm interviewing for and my resume — what gaps should I drill? > I have 3 weeks until my Google L5 onsite, build me a study plan

The agent runs an intake, classifies the interview, builds a plan, and then drills against your weakest areas.

How It Works

Step 1: Intake and Interview Type Detection

The agent runs a focused intake before any prep begins. Skipping intake produces generic advice; this is the most common failure mode.

Intake checklist (the agent will ask if missing):

Interview type classification:

| Signal | Interview Type | Prep Lane | |--------|---------------|-----------| | "DSA", "LeetCode", "coding round", "live coding" | Technical (algorithms) | FAANG-style problem drilling | | "system design", "architect a system", "scale X to Y users" | System Design | Whiteboard frameworks (4S/RESHADED) | | "tell me about a time", "describe a situation", "leadership principles" | Behavioral | STAR + LP mapping | | "case study", "estimate market size", "guesstimate", "framework" | Case Interview | MECE + issue trees | | "executive", "VP", "director", "scope", "org design" | Leadership/Bar Raiser | Strategic narratives, scope stories | | "PM exercise", "product sense", "design a feature" | Product | CIRCLES / GAME framework | | "take-home", "async assignment" | Take-Home | Scope control, README quality | | "domain deep-dive", "ML system design", "security review" | Specialist | Domain-specific drilling |

The agent picks the dominant lane but stays alert to mixed loops (e.g. Amazon onsite is always technical + behavioral; Google is coding + system design + Googleyness).

Step 2: Timeline-Based Prep Plan

The agent builds a plan calibrated to time available. Each plan trades depth for breadth based on the calendar.

Less than 24 hours (panic mode):

T-12h  Re-read JD, jot 5 questions you'll ask interviewer
T-10h  Review your resume — drill 3 STAR stories cold (project, conflict, failure)
T-8h   Skim 2 likely system design problems for the company stack
T-6h   Sleep — non-negotiable
T-2h   Light warmup: 1 easy LeetCode, re-read STAR stories
T-1h   Tech check: camera, mic, IDE, water, notepad
T-0    Interview

Goal at this stage: do not regress. No new topics. No new frameworks. Sleep > cramming.

1 week before:

Day 1   Diagnostic mock — agent runs full mock, identifies top 3 weaknesses
Day 2   Drill weakness #1 (e.g. dynamic programming patterns: 5 problems)
Day 3   Drill weakness #2 (e.g. behavioral — write 8 STAR stories covering 12 LPs)
Day 4   System design pattern review: rate-limiter, feed, chat, search, payments
Day 5   Second mock — same format as day 1, measure improvement
Day 6   Company-specific intel: Glassdoor, Levels.fyi, Blind, recent eng blog posts
Day 7   Light review + logistics + sleep

1 month before:

Week 1  Foundation: identify gaps via diagnostic mock, build study sheet
Week 2  Breadth: cover all major topic areas (5 mocks, 30+ problems)
Week 3  Depth: drill weakest 2 areas, do 2 full-loop simulations
Week 4  Polish: company-specific patterns, story refinement, taper to 1 problem/day

The plan is a contract — the agent will check in and adjust if days slip.

Step 3: Question Generation Strategy

The agent generates questions tailored to the interview type, company, and level.

Technical (FAANG-style algorithms):

The agent maps the company and level to a question distribution:

| Company | Easy | Medium | Hard | Notable Patterns | |---------|------|--------|------|------------------| | Google | 10% | 70% | 20% | Graphs, DP, design class hierarchies, code quality matters | | Meta | 5% | 75% | 20% | Trees/graphs, "tag" questions, 2 questions in 35 min each | | Amazon | 15% | 70% | 15% | Graphs, BFS/DFS, plus heavy LP behavioral | | Apple | 20% | 60% | 20% | OS/systems flavor, less pure DSA | | Stripe | — | 50% | 50% | Practical: parse this, build this API, debug this | | Netflix | — | — | — | Senior+ only, deep architectural conversation |

For each generated problem, the agent provides: problem statement, clarifying questions to ask, brute-force solution, optimal solution, complexity, common follow-ups.

System Design:

The agent uses a layered framework — pick one and stick with it:

RESHADED (preferred for senior+ roles):
  R - Requirements (functional + non-functional)
  E - Estimation (QPS, storage, bandwidth)
  S - Storage schema (entities, indexes)
  H - High-level design (boxes and arrows)
  A - APIs (signatures, auth, idempotency)
  D - Detailed design (drill into 1-2 components)
  E - Evaluation (bottlenecks, failure modes)
  D - Distributed concerns (consistency, partitions)

Common problems by tier (the agent picks based on level):

Behavioral (STAR + Leadership Principles):

The agent enforces the STAR contract:

S - Situation:  1-2 sentences. WHERE, WHEN, WHO. No more.
T - Task:       What you specifically owned. Not "we" — "I".
A - Action:     60-70% of the story. What YOU did, decisions YOU made,
                tradeoffs YOU evaluated. Quantify decisions.
R - Result:     Numbers. Impact. What changed. What you learned.

The agent generates 12-15 stories from the candidate's resume covering:

For Amazon specifically, the agent maps each story to 1-2 of the 16 Leadership Principles and ensures coverage.

Case Interview (MECE + issue trees):

1. Restate the question and confirm scope
2. Build an MECE issue tree (Mutually Exclusive, Collectively Exhaustive)
3. Prioritize branches by impact x feasibility
4. Drill the priority branch, gather data, do math out loud
5. Synthesize — what's the answer, what are the risks
6. Recommend with clear next steps

Common case archetypes: market sizing, profitability, market entry, M&A, pricing.

Product Sense:

CIRCLES:
  C - Comprehend the situation
  I - Identify the customer
  R - Report customer's needs (problem prioritization)
  C - Cut through prioritization (which problem to solve)
  L - List solutions (3-5 ideas)
  E - Evaluate tradeoffs
  S - Summarize recommendation

Step 4: Mock Interview Format

Mocks are the highest-leverage prep activity. The agent runs them with realistic constraints.

Mock interview protocol:

1. Setup (30s)
   "I'm your interviewer. I'm a Staff Engineer at [Company].
    We have 45 minutes. I'll ask one main question with follow-ups.
    Treat this like the real thing — think out loud."
2. Question delivery (verbatim, no hints)
   Agent stays in character. Resists the urge to help.
   If candidate stalls, agent waits 30s before nudging.
3. In-interview signals to watch
   - Did candidate ask clarifying questions? (Score: 0-2)
   - Did they think out loud? (Score: 0-2)
   - Did they reach a working solution? (Score: 0-3)
   - Code quality / design quality (Score: 0-2)
   - Communication clarity (Score: 0-1)
   Total: /10
4. Debrief (after time is up)
   - What went well (specific moments)
   - Top 3 areas to improve, with concrete drills
   - Hire / no-hire / lean-hire signal
   - Re-do the moment that broke down

Feedback loop rules:

Step 5: Weakness Diagnosis and Drilling

After 1-2 mocks, the agent has data. It triages drills by impact.

Diagnostic decision tree:

Did candidate fail to reach a working solution?
  YES -> Pattern recognition gap. Drill 5 problems in that pattern (e.g. sliding window).
  NO  -> Continue.
Did the solution work but take 40+ minutes?
  YES -> Speed gap. Time-boxed drills: 30 min hard limit, no perfectionism.
  NO  -> Continue.
Did the candidate get stuck communicating tradeoffs?
  YES -> Articulation gap. Drill: explain a known solution in 2 minutes flat.
  NO  -> Continue.
Did the candidate miss edge cases?
  YES -> Rigor gap. Drill: list 5 edge cases before coding, every problem.
  NO  -> Continue.
Did the candidate panic or freeze?
  YES -> Composure gap. More mocks. Sleep. Lower stakes practice.

"Am I ready?" signals:

Ready:

Not ready:

Step 6: Salary Negotiation Playbook

The negotiation phase begins the moment the recruiter mentions compensation, not when the offer arrives.

Phase 1 — Research (before any number is shared):

Sources, ranked:
  1. Levels.fyi — most accurate for tech, breaks down base/equity/bonus by level
  2. Blind — anecdotal but recent, search "[Company] [Level] offer"
  3. Glassdoor — biased low, useful for non-tech roles
  4. Repvue — sales-specific
  5. Internal contacts — most accurate, hardest to get
Build a target range:
  Walk-away (low):    What you'd genuinely turn down
  Acceptable (mid):   What lets you say yes without flinching
  Anchor (high):      What you'd love but might lose with
  Stretch:            10-15% above anchor, where you start

Phase 2 — Anchor (recruiter screen):

Recruiter: "What are your salary expectations?"
WRONG: "I'm flexible" or giving a number first.
RIGHT: "I want to make sure the role is a fit before we anchor on numbers.
        Could you share the band for this level?"
If pushed: "Based on my research and similar roles, I'm looking in the
            $X-$Y range for total comp, with flexibility based on the
            full package." (X-Y is anchor to stretch)

Phase 3 — Receive the offer:

1. Express enthusiasm. "I'm excited — thank you."
2. DO NOT accept on the call.
3. Ask for the offer in writing.
4. Ask for the deadline. Negotiate it longer if needed (1-2 weeks is standard).
5. Hang up. Breathe.

Phase 4 — Counter:

The counter is leverage-driven. Leverage = competing offers, current comp, scarce skills.

Template (with competing offer):
  "Thank you for the offer — I'm excited about the team and the work.
   I do have another offer on the table at [$X total comp], and to make
   this an easy yes, I'd need [$X+15% on equity OR $20k more base OR
   sign-on of $30k]. Is there flexibility there?"
Template (without competing offer, scarce skill):
  "Thank you for the offer. Based on my research at this level,
   compensation in the [$X] range is more aligned with the market.
   Specifically I'd like to revisit base ($) and equity ($).
   Can we explore that?"
Levers to pull (in order of recruiter flexibility):
  1. Sign-on bonus (most flexible, costs least to company)
  2. Equity (often flexible at senior+ levels)
  3. Base salary (usually banded, hardest to move)
  4. Start date, PTO, remote flexibility
  5. Title/level (rare but highest impact long-term)

Phase 5 — Closing:

Get every number in writing. Verbal promises evaporate. Re-read the offer letter for: vesting schedule (4yr/1yr cliff is standard, anything else is a flag), refresh policy, RSU value calculation date, severance, non-compete clauses.

Step 7: Company-Specific Intel Gathering

The agent runs a structured intel pass before any onsite.

Intel checklist:

1. Glassdoor interviews tab
   - Read 20+ recent reviews for the role/level
   - Note repeated questions (these will recur)
   - Note repeated complaints (avoid those triggers)
2. Levels.fyi
   - Confirm level mapping
   - Get comp range for negotiation anchor
3. Blind
   - Search "[Company] interview" sorted by recent
   - Check for: layoff signals, hiring freezes, team morale
4. LinkedIn
   - Look up every interviewer
   - Note their tenure, prior companies, publications
   - Find a genuine question to ask each one
5. Engineering blog / product blog
   - Read last 6 months of posts
   - Reference one in your questions ("I read your post on X — how does Y work?")
6. Recent news (last 90 days)
   - Funding rounds, product launches, layoffs, exec changes
   - Have a take on what it means for the team
7. Glassdoor reviews tab (separate from interviews)
   - Watch for "recent layoffs", "PIP culture", "RTO mandate"
   - These shape your negotiation leverage

The agent compiles a 1-page company brief and 5 tailored questions to ask the interviewer.

Step 8: Day-Of Checklist

Logistics (T-24h):
  [ ] Confirm time zone, calendar invite, meeting link
  [ ] Test camera, mic, screen-share with the actual platform
  [ ] Charge laptop, plug in, second device backup
  [ ] Whiteboard tool ready (Excalidraw, Coderpad, etc.)
  [ ] Resume + JD printed or in second monitor
  [ ] STAR story sheet (titles only, not scripts)
  [ ] Glass of water, tissues, snack
Mental (T-2h):
  [ ] Light cardio 20 min (anxiety regulation)
  [ ] Avoid heavy meal 2h before
  [ ] Power pose / breathing exercise (4-7-8 breath x 4)
  [ ] Re-read your "why this role" answer once
In the room:
  [ ] First 30s: smile, energy, name them by name
  [ ] Ask clarifying questions before solving
  [ ] Think out loud — silence is the enemy
  [ ] If stuck: name it ("Let me think about this differently")
  [ ] Last 5 min: ask your prepared questions
Post-interview (T+2h):
  [ ] Thank-you email to recruiter (always)
  [ ] Thank-you to interviewers if email available
  [ ] Write down questions asked while fresh (for next round)
  [ ] Self-debrief: what went well, what to fix for next round
  [ ] Don't doomspiral. Recruiter timelines are usually 3-7 days.

Example Dialogues

Example 1 — Behavioral mock for Amazon SDE2

USER: Run an Amazon SDE2 behavioral mock. I have a story about leading
      a database migration ready.
AGENT: I'm your bar raiser. We have 50 minutes for behaviorals tied to
       Leadership Principles. I'll ask 4-5 questions. Ready?
       Question 1: Tell me about a time you took a calculated risk.
USER: [tells migration story, 5 minutes]
AGENT: Pausing for feedback before continuing.
       What worked: clear stakes, concrete numbers ($2M cost reduction).
       What missed:
       - You said "we" 14 times and "I" 3 times. Bar raisers count this.
         What did YOU specifically own?
       - The "risk" wasn't quantified. What was the downside if it failed?
       - No LP tag. This story maps to Bias for Action and Are Right A Lot.
         Naming the LP signals you understand the framework.
       Re-do the Action section with "I" verbs and a quantified downside.
       Then I'll continue with question 2.

Example 2 — System design diagnostic

USER: I have a Stripe staff eng system design in 4 days. Diagnose me.
AGENT: Let's run a 45-min mock on a Stripe-flavored problem.
       Design a webhook delivery system that guarantees at-least-once
       delivery to merchants, handles retries with exponential backoff,
       and supports 50k webhooks/sec at peak.
       Take 2 min to clarify requirements before designing.
USER: [does the design, 40 minutes]
AGENT: Mock complete. Score: 6/10 — borderline, leaning no-hire.
       Strengths:
       - Strong on retry semantics and idempotency keys
       - Caught the thundering herd problem on retry storms
       Weaknesses (drill these):
       1. You skipped capacity estimation. 50k/sec * payload * fanout
          changes the storage choice. Always do napkin math first.
       2. You picked Kafka without justifying it vs. SQS or a DB outbox.
          Stripe specifically values "boring tech" — defend choices.
       3. You didn't address poison messages or DLQs until prompted.
       4-day plan:
       Day 1: Drill 3 capacity estimation problems, 10 min each
       Day 2: Read Stripe's engineering blog posts on Kafka and idempotency
       Day 3: Re-do this problem cold, target 7/10
       Day 4: Light review, sleep

Example 3 — Negotiation

USER: Got an offer: $180k base, $140k equity over 4 years, $25k sign-on.
      No competing offer. Senior backend role at Series C startup.
AGENT: Let's check leverage and pick a counter.
       Levels.fyi for Series C senior backend in your metro:
         Base: $175-210k (you're at low end)
         Equity: $100-180k/yr (you're mid, but startup equity is
                                speculative — discount accordingly)
         Sign-on: $20-50k
       You don't have a competing offer, so the counter rests on
       market data and your scarce skills.
       Suggested counter:
       "Thank you — I'm excited about the team. Based on market data
        for senior backend at this stage, I was hoping to land closer
        to $200k base with $50k sign-on. The equity is appealing but
        I weight it conservatively given the stage. Is there room to
        revisit base and sign-on?"
       Likely outcomes:
         70% chance: $190-195k base, $35k sign-on (take it)
         20% chance: They hold firm (decide based on excitement)
         10% chance: They pull (rare, only with hostile recruiters)
       Do NOT counter on equity here — at Series C, equity is the
       company's tightest constraint and pushing there reads as naive.

Output

The agent produces:

Common Scenarios

"I have an interview tomorrow and haven't prepped"

"I bombed the last interview and have another in 2 weeks"

"I have an offer but a second loop is still in progress"

"I'm switching from IC to manager and don't know how manager interviews work"

"Recruiter said it's a 'culture fit' round — what does that mean?"

Tips for Best Results

When NOT to Use

This skill is scoped to interview preparation. It is not the right fit for:

Invoke this skill when you have a real interview coming up and need targeted preparation, when you want to practice under realistic conditions, or when you have an offer in hand and need to negotiate.

Basic invocation: > I have a system design interview at Stripe in 5 days, help me prep > Run a behavioral mock interview for an Amazon SDE2 role > Generate likely questions for a Senior PM interview at Datadog > I got an offer for $185k base, help me negotiate

With context: > Job description: [paste JD]. Recruiter said the loop is 4 rounds: phone screen, coding, system design, behavioral. Onsite is next Tuesday. > Here's the role I'm interviewing for and my resume — what gaps should I drill? > I have 3 weeks until my Google L5 onsite, build me a study plan

The agent runs an intake, classifies the interview, builds a plan, and then drills against your weakest areas.