Pillar 2 of 6
2

Making Your Job Description Your Best Sales Page

Tech companies build beautiful product pages and customer acquisition funnels, but forget that talent acquisition needs the same love and care.

TL;DR

Core Insight: The Job Scorecard from Pillar 1 becomes 4 assets that work together to attract and filter the right candidates.

The 4 Assets:

  1. JD Sales Page - The 10-section formula that sells the role (not lists requirements)
  2. Role VSL - 60-90 second video that puts a face to the role
  3. Application Form - Sells while they apply, filters with smart questions
  4. Recruiter Pitch - The talking points for screening calls (knockout question, one-liner, 3 core strengths)

Every asset pulls directly from the Scorecard. One source of truth, consistent messaging everywhere.

The Core Philosophy

  1. Sales page, not requirements list. The JD is selling the best candidates in the world
  2. Every asset pulls from the Scorecard. One source of truth, consistent messaging everywhere
  3. Paint the 18-month outcome. Top performers want to see the scoreboard
  4. Identity statements, not checklists. Write requirements as who they ARE
  5. Personal sign-off matters. Human connection beats corporate polish

How the Scorecard Becomes Your Assets

The Job Scorecard from Pillar 1 feeds directly into every asset you create here:

Scorecard Section→Asset Created
Vision for Role→JD headline, VSL hook
Single Sentence Filter→JD identity statement, outreach messaging
Dream Role Overview→JD success picture, VSL benefits
Required Skills (Max 3)β†’JD requirements, VSL requirements, application questions
Candidate Profile→Application filter questions
Comp Range→JD perks section

The 4 Assets You Create in Pillar 2

AssetPurpose
1. JD Sales PageConvert passive browsers into excited applicants
2. Role VSLPut a face to the role, build trust before they apply
3. Application FormSell while filtering, reveal how candidates think
4. Recruiter PitchConsistent messaging on screening calls, qualify early

The 10-Section Structure

#SectionPurpose
1HookFuture vision question
2IdentityWho they ARE (not what they do)
3CredibilityWhy this company is legit
4Day-to-DayWhat they'll actually work on
5Success Picture18-month outcomes
6PerksReal career benefits (not snacks)
7RequirementsWritten as identity statements
8ValuesReal filters, not corporate fluff
9ProcessClear next steps
10ClosePersonal sign-off from leadership

Each section has a specific purpose. The next steps break down exactly how to write each one.

Section 1: The Hook (Future Vision)

Formula: "Are you a [role] ready to become [outcome]?"

Are you a [role type] ready to become [niche-famous outcome] as [company positioning]?

Imagine it's [future year] and you're the key [role] that transformed [Company] from [current state] to [future state].

Within just a few years, you've [impact statement].

Why it works: Sells the VISION of who they'll become, not the job tasks.


Section 2: Identity Statement

Formula: "You are not a [common label]. You are a [elevated identity]."

But to achieve this, you didn't rely on [common tactics]. Shiny tactics are useless if [contrarian insight].

That's not who you are. You are a [Elevated Identity].

You're not just a "[common label]". You are a [Identity of Action].

You have the ability to [capability 1], [capability 2], and [capability 3].

Why it works: Top performers want to be SEEN. This speaks to their identity.


Section 3: Credibility Markers

Formula: Social proof that makes them think "this is legit"

What Makes Us Great?

You've seen us [credibility marker]. You may have come across [products/content].

[Company] is a [type of organization]. Think about [problem your audience faces]. Most people were never taught how to [solution]. That's where our mission comes in.

We [mission statement] to help [audience] achieve [outcomes].

Will you be the person that helps us [vision]?

Why it works: Proves this isn't a random startup. Real traction. Real impact.


Section 4: Day-to-Day Specifics

Formula: Concrete responsibilities, written as "You'll [action]"

What Will You Be Doing Day to Day?

  • You'll work with [leadership team] to [strategic responsibility]
  • You'll [own/scale/build] our [channel/function] by [how]
  • You'll own all outcomes for [team/function]
  • You'll [unique responsibility that excites them]

Why it works: Removes ambiguity. They know exactly what they're signing up for.


Section 5: Success Picture (18 Months)

Formula: Paint the picture of what winning looks like

What Success Looks Like (18 Months Post-Hire)

  • You've [metric/outcome achieved]
  • You've built [team/system] that [outcome]
  • You're known as the key leader responsible for [achievement]
  • Your team is considered the best in the world at [domain]
  • You've [impact statement]
  • You're credited with [legacy/movement created]

Why it works: Top performers are outcome-oriented. Show them the scoreboard.

Section 6: Perks (Real Benefits)

Formula: Career benefits, not office perks

PerkDescription
Development[Skill growth opportunity]
Trajectory[Career positioning benefit]
Flexibility[Work arrangement]
Impact[Mission-level benefit]
Ownership[Autonomy statement]
Mentorship[Who they'll learn from]
Compensation[Honest comp positioning]

Why it works: Nobody cares about ping pong tables. They care about growth.


Section 7: Requirements (As Identity)

Formula: "You've done X" not "Must have X years"

You Will Succeed Here If:

  • This isn't your first rodeo. [Proven track record statement]
  • You've built [thing] before. [Specific experience as identity]
  • You understand [domain]. [Knowledge as capability]
  • You possess [quality]: [examples of that quality in action]
  • You know how to [outcome]. [Outcome-oriented requirement]

If you don't fit 100%, but still think you'd be great for the role, let us know in the application!

Why it works: Reads as "this is who you ARE" not "this is our checklist."


Section 8: Values (Real Filters)

Formula: "X is on brand. Y is off brand."

What We Care About (Our Values)

  • [Value 1] is on brand. [Anti-value 1] is off brand.
  • [Value 2] is on brand. [Anti-value 2] is off brand.
  • [Principle 1]
  • [Principle 2]
  • [Core operating principle]

Why it works: Real values filter out bad fits. Corporate values attract everyone.


Section 9: Clear Process

Formula: Remove all mystery from what happens next

What You Can Expect After Applying

  1. [Step 1] - [who/what/duration]
  2. [Step 2] - [who/what/duration]
  3. [Step 3] - [what they'll do]
  4. [Step 4] - [final evaluation]
  5. [Step 5] - [offer/terms]
  6. Joining! [Onboarding preview]

Why it works: Uncertainty kills applications. Clarity increases conversion.


Section 10: Personal Close

Formula: CEO/Founder personal sign-off

I'm excited that you took the time to read, apply, and are interested in learning more about our mission.

Can't wait to see what you're capable of - [Name], [Title]

Why it works: Human connection. This isn't HR-it's leadership reaching out.

Role VSL (Video Sales Letter)

This becomes a VIDEO on the actual job page. Almost no company does this. It's a massive talent branding differentiator.


The Two-Video Strategy

Every job page should have TWO videos:

VideoPurposeLengthSpeaker
General Recruitment VideoSell the COMPANY (culture, values, team)2-3 minCEO + Team testimonials
Role VSLSell THIS specific role60-90 secCEO or Hiring Manager

How the VSL Pulls from the Scorecard

Scorecard SectionVSL Section
Vision for RoleHook
Dream Role Overview"If you're someone who wants..."
Required Skills (Max 3)"But to be a fit you must..."
Company ContextMission/Backstory

VSL Script Structure

Section 1: Hook (Grab Interest)

Are you an incredibly talented [role]?

[Insert vision for role from scorecard]

If so, the [role name] at [Company] may be a fit for you.

Section 2: Company Vision & Mission

[Insert compelling industry problem/stat]

And that brings us to the mission of [Company].

We're a [type of company] dedicated to [mission statement].

Over the last [time period], we've [impressive traction].

But the most exciting thing? At [Company], we are just getting warmed up.

Section 3: "If You're Someone Who Wants..." (Benefits)

And in order to achieve this mission - we need YOU as a [role name].

If you're someone who wants:

  • [Dream element 1 from scorecard]
  • [Dream element 2 from scorecard]
  • [Dream element 3 from scorecard]

Section 4: "But To Be A Fit You Must..." (Requirements)

But in order to be a fit you must:

  • [Required skill 1 from scorecard]
  • [Required skill 2 from scorecard]
  • [Required skill 3 from scorecard]

Section 5: The Honest Warning + CTA

Now, fair warning - being on the team at [Company] is absolutely not for everyone.

I can't promise it will be easy, but I can promise you'll work with some of the best people in the world and feel rare fulfillment by directly [impact statement].

So consider applying. I would love to talk with you myself personally.


Production Notes

ElementRecommendation
SpeakerCEO/Founder or Hiring Manager
Length60-90 seconds
StyleDirect to camera, authentic, not over-produced
PlacementEmbedded on job page, above the fold

Wispr Flow Application: ML Engineer VSL Script

Speaker: Tanay Kothari (CEO) or Sahaj Garg (CTO)

Length: ~90 seconds

This is what the VSL template looks like when filled out for the ML Engineer role. This video would be embedded on the job page.


THE SCRIPT

Section 1: Hook

Are you an ML engineer who's tired of optimizing metrics that don't matter?

Who wants to ship code that millions of people use every single day?

Who obsesses over latency because you know 50 milliseconds is the difference between magic and frustration?

If so, the ML Engineer role at Wispr might be the best opportunity of your career.


Section 2: Mission & Backstory

Here's the thing most people don't realize:

People type at 40 words per minute. But they think at 400 words per minute.

The keyboard - this thing we've used for 150 years - is a bottleneck to human potential.

That's why we built Wispr Flow.

We're on a mission to replace the keyboard entirely. To make talking to your computer feel as natural as talking to a friend.

We've raised $81 million. We're growing 50% month over month. And we're the first voice product that people actually use MORE than their keyboards.

But we're just getting started.


Section 3: "If You're Someone Who Wants..."

And to get where we're going, we need you.

If you're someone who wants:

  • Ownership, not org charts - At 15 people, you'll own entire systems. Not a slice of a slice.

  • Work that ships to millions - Your code runs every time someone talks to their computer. You'll see users tweet about your features.

  • A team that chose this over Google and Anthropic - We're small, elite, and moving fast. No bureaucracy. Ship Monday, see data Tuesday.


Section 4: "But To Be A Fit..."

But to be a fit for this role, you need to:

  • Have shipped ML to production - Real systems with real users, not just notebooks.

  • Obsess over latency - You know the difference between P50 and P99, and you care deeply about both.

  • Move fast without breaking what matters - We're a startup. We need people who can debug across the entire stack.


Section 5: Honest Warning + CTA

Fair warning - Wispr is not for everyone.

I can't promise it'll be easy. We're building something that's never existed before.

But I can promise you'll work on problems that matter. You'll learn faster than anywhere else. And in 10 years, you'll look back and say "I helped build that."

If that sounds like you, apply now.

I'd love to meet you.


Production Notes

ElementRecommendation
SettingOffice or home office, casual but clean
ToneAuthentic, direct, not overly polished
Eye contactLook at camera (teleprompter helps)
WardrobeWhat you'd actually wear to work
B-roll optionsProduct demo clips, team working, user reactions

How This Video Gets Used

ChannelUsage
JD PageEmbedded above the fold
LinkedInPosted with transcript snippets
Cold Outreach"Here's our CEO talking about the role"
Recruiter Screen"Did you watch the video? Thoughts?"

Application Form Optimization

Job applications are absolutely broken. The application form is a great selling opportunity and filter for people who are obsessed about Wispr Flow.


The Problem with Most Application Forms

Standard ATS FormsResult
Feel like bureaucracyCandidates feel nothing
Ask for resume + cover letterBoring, no differentiation
Give no insight into cultureNo self-selection
Don't filter for fitLow-quality applications
Feel like applying to a black holeHigh drop-off

The Better Way: Application as Sales Page

Your application form should:

  1. Reinforce the opportunity (they feel excited while applying)
  2. Filter for fit (wrong people self-select out)
  3. Reveal candidate quality (questions that show how they think)
  4. Feel different (pattern interrupt from every other application)

Form Structure Overview

#SectionPurpose
1The HookReinforce excitement before questions
2The BasicsQuick wins, build momentum
3Filter QuestionsReveal how candidates think (2-3 max)
4Work SamplesResume + portfolio uploads
5The Close + NurturingSet expectations, then keep selling for 5-7 days

Section 1: The Hook

Don't start with "Upload Resume." Start with reinforcement:

You're applying to [Role] at [Company].

We're building [one-sentence mission].

This role is for someone who [ideal candidate description].

If that's you, let's go.


Section 2: The Basics

FieldRequired
Full NameYes
EmailYes
LinkedIn URLYes
Portfolio/GitHubIf relevant

Section 3: Filter Questions (2-3 max)

Question TypeExampleFilters Out
"Why Us"Why [Company]? What specifically resonates?Mass appliers who didn't research
"How You Think"Describe a time you [relevant challenge]. What did you do?Reveals problem-solving approach
"Self-Awareness"What are you actively working to improve?Reveals growth mindset

Section 4: Work Samples

FieldRequired
Resume uploadYes
Portfolio linkOptional
Anything else you want us to see?Optional

Section 5: The Close + Nurturing Sequence

Don't just say "Thanks, we'll be in touch." Keep selling.

Day 0: The Auto-Confirmation (Immediately After Submission)

Subject: We received your application!

Hey [First Name],

Thank you! We have received your application to join [Company].

We will be reviewing your answers in detail... then I'll be in touch with next steps.

In the meantime, check out this video on YouTube from our CEO, [CEO Name].

[1-2 sentence company mission/social proof]

I hope you're as excited about it all as we are. It's definitely going to be a fun ride. :)

Stay tuned for an email from me as soon as we review your application.

Talk soon, [Recruiter Name]

But that's just Day 0. This person is excited. They took time out of their day to apply. And what happens next? Black box. They never hear back unless someone actually reviews their resume. Which almost no one does because resumes don't really work anyway.

You just collected their email. Use it.

The 5-7 Day Nurturing Sequence runs automatically regardless of whether they get the job:

DayEmailPurpose
0Application receivedConfirmation + what to expect
1Company storyWhy we exist, the problem we're solving
2Team spotlightWho they'd work with, culture proof
3Product deep diveWhat we're building, recent wins
4Customer/user storiesImpact we're having
5Behind the scenesDay in the life, office culture
6-7Final touchReminder they matter, next steps

Why this works:

BenefitImpact
Candidates stay warmEven if process takes weeks, they're engaged
Builds employer brandThey tell friends about the experience
Creates advocatesRejected candidates still love the company
Reduces ghostingThey feel invested, less likely to drop
Pre-sells the offerBy the time you extend, they're already sold

People know it's automated. They don't care. They get to learn more about the company they might join. That's valuable.


Example: ML Engineer Application

Section 1: The Hook

You're applying to ML Engineer at Wispr.

We're replacing the keyboard. Building voice interfaces so good that talking to your computer feels like talking to a friend.

This role is for someone who ships production ML systems, obsesses over latency, and wants their code used by millions.

If that's you, let's go.


Section 2: The Basics

FieldRequired
Full NameYes
EmailYes
Link to an interesting piece of work you're proud ofOptional
GitHubOptional

Section 3: Filter Questions

TypeQuestion
"Why Us"Why Wispr? What about our mission made you want to apply?
"How You Think"Tell us about a time you significantly improved latency or performance of an ML system in production.
"Self-Awareness"What's one area of ML engineering you're actively working to improve?

Section 4: Work Samples

FieldRequired
ResumeYes
Anything else you want us to see?Optional

Section 5: Candidate enters the nurturing sequence (Day 0 confirmation + 5-7 day drip)


Common Mistakes

MistakeFix
Too many required questions3 custom questions max
Asking for cover letterReplace with specific questions
Generic "Tell us about yourself"Ask role-relevant questions
No confirmation messageTell them exactly what happens next
Same form for every roleCustomize key questions per role

Recruiter Pitch Template

The pitch is how you SELL the role on calls. It's not the JD. It's the conversation. Every recruiter should have this document memorized before screening candidates.


The Philosophy: Interview to Sell, Not to Assess

This pitch connects directly to Pillar 4: Interview to Sell. The whole idea is you're interviewing to SELL, not to assess.

Don't pretend to be more technical than you are. Wispr Flow has incredible engineers who can run the technical assessments. That's not your job.

Your job is to know enough to tell the difference between average, good, and great. You need to be technical enough to understand the signal, not to evaluate it yourself.

What This MeansWhat It Looks Like
Be obsessed with Wispr Flow's productUse it daily. Know every feature. Understand why it matters
Learn to vibe codeBuild internal tools that help you hire incredible people. Build demos. Make your ideas come to life
Become 100x betterVibe coding turns you from a recruiter into a recruiting engineer. You can build what you imagine
Learn constantlyRead the papers. Understand the basics of ML. Know what latency means
Stay in your lane on technicalsGet the signal, then pass to engineers for deep assessment

The goal: You should be dangerous enough to qualify, but humble enough to know you're not the final judge. Your conviction should come from understanding the mission and selling it authentically, not from pretending to know more than you do.


How to Use This Pitch

This should be a living document. Create a separate page or doc for each role's pitch. Link to it. Reference it before every call.

What a good recruiter does is adjust the pitch based on how the interviews go. What questions are people asking? What's consistent? You want to constantly update and refine this pitch to make it better and better and better.

Pay Attention ToWhat To Do
People light up when you say certain thingsMark it. That's gold. Use it more
People get bored when you say other thingsCut it. Ruthlessly
Questions that keep coming upAdd answers to your pitch proactively

Review your pitches like a professional athlete reviews their game tape. The goal is to improve. To get better. Because the stakes are high.

People are potentially leaving their current opportunity, their job, to work on your team. That's a huge commitment. So you want to show up and be enthusiastic. Give it your absolute full for the pitch.


How This Connects to Everything Else

The Recruiter Pitch is Asset 4. It pulls directly from everything you've already built:

SourceWhat You Pull
Job Scorecard (Pillar 1)Vision for Role β†’ Knockout Question, One-Liner
Job Scorecard (Pillar 1)Required Skills (Max 3) β†’ 3 Core Strengths
Job Scorecard (Pillar 1)Single Sentence Filter β†’ Intro Call Requirements
JD Sales Page (Asset 1)Hook β†’ Mission Hook
JD Sales Page (Asset 1)Day-to-Day β†’ Role Scope
JD Sales Page (Asset 1)Success Picture β†’ Growth & Opportunity
Application Form (Asset 3)Filter Questions β†’ Questions to Ask

The pitch is not new content. It's a restructuring of everything you've already created for a live conversation.


The Philosophy: Seeing People Clearly

This section is inspired by Graham Duncan's essay "What's Going On Here", one of the best pieces ever written on screening people.

The Rider and the Elephant

When you interview someone, visualize two elephants in the room: yours and theirs. Your conscious mind (the rider) is talking to their conscious mind (their rider). But the elephants, the unconscious drives, are what actually determine behavior.

What You SeeWhat's Actually Happening
What they SAY they'll do"Espoused theory"
What they ACTUALLY do"Theory in use"

Your job is to see BOTH. The rider is easy. The elephant requires better questions.

The Goal

Graham Duncan describes the ideal: "If the way Steve Jobs and Jony Ive knew each other after 14 years of daily conversation was a 10, how close to that 10 can you get when you're first meeting someone?"

Most people start at a 3. With the right process, you can get to a 7.


Pitch Structure Overview

#SectionPurposePulls From
1IntroPersonal, disarming, lower stakesApplication answers
2Knockout QuestionONE filter to qualify earlyScorecard: Vision for Role
3Top 2 RequirementsConfirm fit before going deepScorecard: Required Skills
4One-LinerWhat you do in one sentenceScorecard: Vision
5The ProblemPain you solveJD: Hook
6The SolutionYour product, transformationJD: Credibility
7The Role ScopeWhat they'll ownJD: Day-to-Day
83 Core StrengthsIdeal candidate profileScorecard: Required Skills
9Mission HookWhy beyond moneyJD: Success Picture
10Growth & OpportunityWhere company is goingJD: Perks
11Questions to AskSee the elephantGraham Duncan framework
12Development & BenefitsWhat they'll becomeJD: Success Picture

1. Intro

Say something personal. Reference their background, what stood out in their application. Lower the stakes, put them at ease.

What to DoWhy
Reference their applicationUse the filter questions you asked in Asset 3
Mention what caught your attentionShows you actually read their submission
Make them feel seenBefore you start evaluating

Goal: Get them relaxed enough to show you the elephant, not just the polished rider.


2. Knockout Question

Before going deep, have ONE filter question that determines if they're even in the running.

Pull from Scorecard: Your "Vision for Role" defines what this person must have done.

Format: "Have they [specific experience/achievement]? If not, they will not be a likely fit."

Why this matters: Don't waste 45 minutes on someone who doesn't clear the bar. Qualify in the first 5 minutes.


3. Intro Call Top 2 Requirements

#RequirementPulls From
1[Specific skill/experience check]Scorecard: Required Skills
2Does [Company] align with what they're looking for?Always ask this

Why "alignment" matters: Even if they're qualified, if your mission doesn't resonate, they won't last. Better to find out now.


4. One-Liner

Pull from Scorecard: Your "Vision for Role" in one sentence.

Format: "We [what you do] for [who you serve] through [how you do it]."


5. The Problem

Pull from JD: This is your Hook section, restructured for conversation.

QuestionAnswer
Who suffers from this problem?[Target audience]
What does their life look like now?[Current pain]
Why hasn't it been solved?[Gap in market]
What's at stake if nothing changes?[Consequences]

6. The Solution

Pull from JD: This is your Credibility section, restructured for conversation.

ElementDescription
What you sell[Product/service]
How it works[Mechanism]
Who delivers it[Team/tech]
Transformation[Before state] β†’ [After state]

7. The Role Scope

Pull from JD: This is your Day-to-Day section.

ElementDescription
Core responsibility 1[What they own]
Core responsibility 2[What they build]
Core responsibility 3[What they lead]
Reports to[Who]
Team size[Number they'll manage/work with]

8. Ideal Candidate Profile (3 Core Strengths)

Pull from Scorecard: Your "Required Skills (Max 3)" become the 3 Core Strengths.

This person has 3 core strengths which align with [Company]:

StrengthWhat This Looks Like
1. [Strength][Specific behaviors/achievements]
2. [Strength][Specific behaviors/achievements]
3. [Strength][Specific behaviors/achievements]

9. Mission & Vision Hook

Pull from JD: This is your Success Picture, framed as identity.

QuestionAnswer
Why would someone want this beyond money?[Mission]
Who will they impact?[Audience/scale]
What will they become known for?[Legacy/identity]

Key Reason: [Type of person] are builders. The person that will want this challenge craves [deeper motivation].


10. Growth & Opportunity

Pull from JD: This is your Perks section, focused on trajectory.

ElementDescription
Financial picture[Profitable? Revenue? Funding?]
Market opportunity[Size, trajectory, position]
Growth path for them[What they'll become known for]

11. Questions to Ask: The Graham Duncan Framework

These questions come from Graham Duncan's essay "What's Going On Here". The goal is to see the ELEPHANT, not just the rider.

How to Use These Questions

PrincipleHow to Apply
Talk very littleCreate stillness. Let them fill the silence
Write down their questionsQuestions have high signal value. Ask "why did you ask that?"
Look for the hungry mindHow their questions flow is hard to fake
Hold negative capabilityStay in partial knowing. If you're not confused at times, you're doing it wrong

See the Elephant (Reveal the Real Person)

QuestionWhat It Reveals
What criteria would you use to hire someone for this job if you were in my seat?How they think about the role, jargon vs original thought
How would your spouse/sibling describe you with ten adjectives?Self-awareness, ability to take another's perspective (works ~50% of time, more with senior candidates)
What are you compulsive about?The elephant, their true drivers
Where have you experienced ignition, seeing an older person doing something and intuiting you were wired the same way?Their identity, what they're drawn to

Future Failure Analysis

QuestionWhat It Reveals
Let's imagine it's 6 months and it didn't work. What's your best guess about what went wrong?Self-awareness about weaknesses, honesty
What are you most torn about right now professionally?Current tensions, what they're wrestling with

Closing

  1. What are you optimizing for in your career now?
  2. What would make this incredibly compelling for you?
  3. How do you feel this interview is going?

12. Development & Benefits

Pull from JD: This is your Success Picture, framed as what they GET.

ElementDescription
Access[What/who they get access to]
Learning[Skills they'll develop]
Become[Identity transformation]

Wispr Flow: ML Engineer Screening Call Script

This is the actual call. Read this before every screening. Memorize the flow. Adjust based on what lights people up.


THE CALL FLOW

The Purpose of This Call:

You're not making a hiring decision on this call. The call is not for that. Unless it's truly someone with proof of work that is so incredible, and there's a different process for that.

The whole point is to excite the person, represent Wispr Flow and the team in the best way possible, and move them towards the next part of the process. The candidate should want to continue to the next call. That's the goal of every session.

How deep the recruiter screen goes should be decided with Tanay, Sahaj, and the team. It shouldn't go too technical. It should be fun and leave a very positive impression.


PHASE 1: OPEN

What You Say:

"Hey [Name]. Great to meet you. Before we get started, I just wanted to say I read through your application, and [specific thing that stood out]. I specifically liked that you said [the thing they said that relates back to Wispr Flow]."

"This call is going to be chill, low pressure. I just want to learn more about your background, share what we're doing here at Wispr Flow, and see if there's a fit."

"I also think about these calls in a different way. My job isn't to try to catch someone or assess their entire skill set in a 30 minute call, which is actually impossible. But I think that you're incredible, and that's why we're on the call in the first place."

"I use the frame: 'What's the best seat for this person?' Whether it's at Wispr Flow or somewhere we're connected to. That's what I'm obsessed with. Putting people in positive feedback loops so that where they spend a majority of their time is actually meaningful to them."

Why This Works: You're lowering the stakes. You're making them feel seen. You're also showing them you're not playing games. This isn't an interrogation. They relax. The real person comes out.


PHASE 1.5: LEARN WHAT THEY KNOW + FILL IN THE GAPS

The Key Insight:

You want candidates to talk more than you. Aim for a 65/35 split. Keep the conversation flowing and make it as productive and fun for them as possible, especially on the recruiter call.

Regardless of whether this person is an exact fit, we're still deciding. People learn over time. People get better. People know people. You want to promote the company and be excited about what Wispr Flow is building. You want to give every candidate a personalized, incredible experience that makes them not hate recruiters.

What You Say:

"Before I get into the role, I'm curious. What do you already know about Wispr Flow? What have you seen online?"

Listen. This tells you what to skip and what to explain.

"Cool, let me fill in the gaps..."


Filling In The Gaps (Use What They Don't Know)

You have a menu of things to share. Pick based on what they already know.

The One-Liner:

"We're replacing the keyboard. Building voice interfaces so good that talking to your computer feels like talking to a friend."

The Problem:

"Here's the thing most people don't realize. People type at 40 words per minute. But they think at 400 words per minute. The keyboard, this thing we've used for 150 years, is a bottleneck to human potential."

"Voice interfaces have sucked for 30 years. Slow, inaccurate, awkward. Most 'AI voice' is cloud-dependent wrappers. Nobody's actually solved this."

The Solution:

"That's what we're doing at Wispr. We built Flow. It's the first voice dictation platform that people actually use MORE than their keyboards. On-device ML. Sub-500ms latency. Context-aware, so it knows if you're in Slack versus email. The goal is zero edit rate."

Social Proof:

"We've raised $81 million. Latest round was $25M led by Notable Capital and Steven Bartlett's Flight Fund. We're growing 50% month over month. Team of 15, scaling to 50."

5 Surprising Things (Deep Cuts)

FactWhy It's Surprising
They killed their own hardware to save the missionThey were building premium AirPods competitors. Investors loved it. They killed it because software reaches billions faster. That's ruthless pragmatism
Tanay has been obsessed with J.A.R.V.I.S. since age 10Watched Iron Man in 2008, pulled his first all-nighter that night to learn to code. This isn't a business opportunity for him. It's a 15-year obsession
They have a Zero Data Retention agreement with OpenAIEven when using GPT-4o, OpenAI is contractually obligated not to train on Wispr data. That's how they landed law firms and enterprise customers
They built the only AI that fluently speaks HinglishThe language of 1 billion people. OpenAI didn't bother. Wispr did. That's the market they're going after
Internal build v90+ before the world saw v1They iterate 10x faster than Big Tech. Every employee uses Wispr to write code, emails, and Slack messages. If you don't use the product, you can't work there

The Arc:

"Act 1 was dictation. We nailed it. Users literally hide Wispr from their coworkers because it's an unfair advantage."

"Act 2 is the action layer. Voice that doesn't just transcribe, it DOES. Schedule this meeting. Write this code. Search my emails."

"Act 3 is J.A.R.V.I.S. The AI companion Tanay dreamed about since he was 10."


Check In:

"Does any of that resonate with you?"

Listen. What they say here tells you a lot. If they're asking questions, good sign. If they're quiet, probe.


PHASE 2: KNOCKOUT + DIVE DEEP

What You Say:

"So what we're looking for in this role... and there's many things to look for in this role... but very much the key thing is someone who has shipped on-device speech recognition with sub-500ms latency to production users. What's your experience around that?"

PATH A: YES (Continue)

"I'd love to learn more. Can you explain how you did that step by step, in the simplest way possible? How would you explain it to someone like my mom who isn't in tech and needs my help to restart an app on her iPhone?"

This does two things: it shows genuine curiosity, and it tests if they actually understand it deeply. If they can't explain it simply, they might not have owned it. Listen for specifics. Did they own it or were they a cog? If they light up teaching you, that's signal.

Optional: Probe Deeper (If Decided With Team)

If the team wants the recruiter screen to go a bit deeper, you can use these questions. They separate "I've worked on ML" from "I've shipped ML to production under constraints."

QuestionWhat You're Looking For
"Walk me through how you'd optimize inference for a model that's hitting 800ms but needs to be under 500ms"Systematic thinking: profiling, quantization, distillation, architecture changes
"How do you think about P50 vs P99 latency?"Understanding of tail latency. Voice is P99-sensitive
"Tell me about a time you debugged something that worked in dev but failed in production"Real production experience. Specifics matter
"How do you know when a model is 'good enough' to ship?"Product sense, not just ML sense

PATH B: NO (Dive Deeper)

Be honest with them and repeat back what you understood.

"Okay, based on my understanding, it sounds like you don't have on-device machine learning experience. But what you said you do have is [repeat back what they said]."

Let them confirm if you understood correctly. Have a conversation around it. Then ask follow-up questions to understand what the ideal best seat is for them:

"What's something you want to build next? Something that when you ship it, you'd say 'Wow, that's impressive. I'm very proud of that.'"

This tells you where they're headed, what excites them, and whether there's a different seat at Wispr or somewhere you're connected to that might be a better fit.

Always give them the pitch. They know people. They might apply later. Leave them excited about Wispr even if it's not a fit for this specific role.


Cheat Sheet: The 6 Challenges at Wispr Flow

Use this to have a back-and-forth conversation. Don't go too deep technically. See what lights them up, what experience they have, what they want to work on.

ChallengeWhat It IsWhat To Say
The Latency ProblemSub-500ms from speech to text on screen"Every millisecond you shave, users FEEL it. Users are mid-thought. Latency breaks their flow."
Accuracy vs SpeedBigger model = accurate but slow. We need both"Can't brute force with GPUs. Have to be clever. Distillation, quantization, architecture. It's a puzzle."
Context AwarenessKnow if user is in Slack (casual) vs Email (formal)"Not a dumb transcription tool. Building something that UNDERSTANDS. Real AI."
Personalization at ScaleEvery user has different accent, vocabulary, style. Model adapts over time"Learning from millions of users while respecting privacy. Everything on-device."
On-Device PrivacyAudio never leaves device. No cloud GPUs to lean on"Privacy isn't a checkbox. It's a constraint that forces 10x better engineering."
The Last Mile99% to 99.9% is brutal. Rare accents, noise, jargon, mumbling"Anyone can build a demo. We're building something that works for everyone, everywhere, every time."

As you share these challenges, watch what lights them up. They'll tell you what they have experience with. That's the signal.


PHASE 3: THE ROLE

What You Say:

"So the role. You'd be working directly with Sahaj, our CTO, and Tanay, the CEO. ML team is 3 people right now, scaling to 10. You'd help shape who we hire."

"The core problems you'd own:"

"One: the latency problem. Sub-500ms from speech to text on screen. Every millisecond you shave, users feel it."

"Two: accuracy versus speed. Bigger model means more accurate but slower. You'd get clever with distillation, quantization, architecture."

"Three: personalization at scale. Models that adapt to each user over time. Learning from millions while respecting privacy. Everything on-device."

"Four: the last mile. 99% to 99.9% is brutal. Edge cases. Rare accents, background noise, jargon, mumbling. Anyone can build a demo. We're building something that works for everyone, everywhere, every time."

The Hook:

"If you do this well, in 18 months you'll be known as the person who made Flow fast. Your code will run every time someone talks to their computer."


PHASE 4: QUALIFICATION QUESTIONS

Graham Duncan Questions (Pick 2-3):

"Let me ask you something. If you were in my seat, what criteria would you use to hire someone for this role?"

Write down their answer. High signal. Are they thinking originally or using jargon?

"What are you compulsive about?"

This reveals the elephant. What they can't help but do. What they obsess over even when no one is watching.

Future Failure Analysis:

"Let's imagine it's 6 months from now and it didn't work out. What's your best guess about what went wrong?"

If they can't answer this honestly, red flag. Self-awareness matters.


PHASE 5: THEIR QUESTIONS

What You Say:

"What questions do you have for me?"

Keep the conversation flowing like you're talking to a good friend. You're both trying to figure out what Wispr Flow is, how it relates to their dreams, and if there's mutual interest to continue.

Common Questions & Answers:

QuestionYour Answer
"What's the tech stack?""Python, PyTorch, on-device inference. But honestly the stack matters less than the problems. We'll use whatever works"
"What's the team like?""15 people who chose this over Google, Meta, Anthropic offers. Small, elite, moving fast. No bureaucracy"
"What's comp?""$130K-$240K base plus generous equity. We're early, the equity matters"
"Remote or in-person?"[Answer based on current policy]
"What happened to the last person in this role?""We're building the team from scratch. You'd be early"

PHASE 6: END ON A HIGH NOTE

People remember the start and the end way more than the middle of conversations. This is where you ease them if they're a great candidate and guide them on literally what the next steps are.

For Strong Candidates:

"I really enjoyed this conversation. We move fast here. Moving forward, my role becomes more of your personal assistant throughout this process. I want to make sure that in between any calls you have with the team, you can reach out directly to me. I'll respond as quickly as I can with questions, assets you may need, more deep cuts if you want them."

Map Out The Process:

"So the next call would be with Sahaj, our CTO. That's more technical, real problems, not leetcode. After that would be [next step]. Then [final step]. Here's what that looks like..."

StepWhat It Is
1. This callRecruiter screen (done)
2. Technical deep diveWith Sahaj (CTO)
3. System design / Discovery projectReal problems
4. Team dayMeet the people you'd work with
5. OfferFast decisions, transparent comp

The Notes Email:

"After this call, I've been taking some notes. I want to make sure I represent your experience and our conversation correctly to the team. So before I send my notes, I'll send you what I have and what I understand from our conversation. If you have a few minutes, please review that email and let me know if I need to make any corrections if I misunderstood anything about your tech background or what you said. Does that sound okay?"

This does two things: it makes them feel respected, and it gives you a chance to correct any mistakes before the team sees your notes.

For Maybe Candidates:

Be honest. It's better to be upfront than pretend, which is why people really don't like recruiters.

"I want to be honest with you. I'm not 100% sure because of [this thing] and [this thing] that you said. Can we go over those in more depth? Maybe I just don't understand."

This gives them a chance to clarify. And it shows you're not playing games.

For Candidates Who Aren't A Fit:

"I really appreciate your time. Based on what we're looking for right now, I don't think this specific role is the right fit. But I want to say something. I believe everyone can learn and improve. I may not even be 100% correct in my assessment."

"Is there any way I can help in your search? I've been obsessed with helping people land their dream jobs and stand out from the resume black box that is HR and recruiting. I also run events and I help connect people. That's what I do."

"If you go through Tanay's connections or Sahaj's connections on LinkedIn and you see that we're connected to someone you'd love to speak to, I'll try my best to make the introduction. Just let me know."

Leave them with something valuable. They'll remember how you made them feel.


KNOW THESE NUMBERS COLD

MetricValue
Total funding$81M
Latest round$25M (Notable Capital + Steven Bartlett's Flight Fund)
MoM growth50% revenue
Current team15 people
Scaling to50 people
Comp range$130K-$240K + generous equity
Product milestoneFirst voice dictation platform people use MORE than keyboards
Steven BartlettPartnership with The Diary of a CEO
Latency targetSub-500ms
Accuracy targetZero edit rate

THE ONE LINE THAT CLOSES

When you need to land the pitch:

"Your code will run every time someone talks to their computer. You'll work directly with Tanay and Sahaj, ship to millions of users, and help build something that fundamentally changes how humans use technology. This is a building role, not a maintenance role. If you're tired of incremental improvements at big companies, if you want your work to ship and matter, this is it."


RED FLAGS TO WATCH FOR

Red FlagWhat It Means
Can't name specific latency numbersHasn't worked on real production systems
Talks only about model accuracy, never user experienceOptimizes for metrics, not users
Blames others for past failuresLow ownership
Gets defensive when you probe on weaknessesPoor self-awareness
No questions for youNot genuinely curious
Asks about WFH/hours before asking about the productOptimizing for lifestyle, not mission
"I've done a lot of X" without specificsProbably didn't do X
Can't explain technical concepts simplyDoesn't deeply understand them

HANDLING OBJECTIONS

The Two Types of Objections

TypeExamplesHow to Handle
Logistics-based"I can't leave my job for two weeks." "The salary is too low."Easy. These are logical. Solve the logistics.
Fear-based"I don't think you're the right fit." "I'm not sure you're as good as you say."Hard. They don't trust you yet. Have an honest conversation and help them gain clarity.

Key insight: Many logistics objections are actually fear-based objections in disguise. "The money isn't right" often means "I'm not convinced this is worth the risk." Dig deeper before trying to solve the surface problem.


The Goal: Meeting of the Minds

Before any contracts, you need a meeting of the minds. An agreement between two people that this is the right fit.

These examples assume the candidate is a good fit, they agree Wispr Flow is right for them, and now you just need to collaborate to create a scenario both parties are happy with. The exact playbook for each objection stays internal to Wispr Flow. These are high-level frameworks for hypothetical scenarios.


The Framework

StepWhat It Means
1. Uncover the real objectionIs it logistics or fear? Nothing else works until you know which one you're solving.
2. Acknowledge firstDon't be defensive. Validate their concern.
3. Pivot to the real valueReframe around what actually matters to them.
4. Never trash the competition"OpenAI is a great company. Different stage, different tradeoffs."
5. Be honest about the stageDon't oversell stability. Be real about what startup life means.

Example Objections

ObjectionResponse
"OpenAI offered $450K""At OpenAI, you'd be engineer #847. Here, you'd be one of the first. The system you build will BE the product. We can't match Big Tech cash, but the equity window is still wide open."
"You're using Whisper like everyone else""We build on Whisper, but we run on-device. No cloud to fall back on. At Google you solve problems with scale. Here you solve them with engineering. The demo is easy. Making it work everywhere, every time? That's the puzzle."
"I'm worried about startup risk""$81M raised. 50% month-over-month growth. Steven Bartlett as an investor. We're not pre-product hoping something works. We have something that works, and we're scaling it."
"I've never worked on-device before""What matters is whether you're excited to learn. If sub-500ms latency without cloud GPUs sounds like a puzzle, we can teach you. If it sounds like a headache, probably not the role."
"I need to think about it""What specifically? Is it the role, the comp, the stage, or something else?" Don't let them leave without understanding the blocker.
"I want to do research, not engineering""We publish papers. But more importantly, you'll ship research to millions of users. That's rarer than you think."

Things to Never Do:

NeverWhy
Make up technical specificsEngineers will catch you. Say "I don't know, but I can find out"
Trash competition"OpenAI is a great company. Different stage, different tradeoffs"
Get defensive about constraintsFrame them as what makes it interesting
Oversell stabilityBe honest about the stage
Promise what you can't deliverBe honest about comp, scope, timeline

AFTER THE CALL

Do ThisWhy
Write notes immediatelyYou'll forget details in 30 minutes
Rate them 1-10 on the 3 core strengthsLatency obsession, ownership, user empathy
Note what lit them upUse this in future pitches
Note what fell flatCut this from future pitches

Quick Reference: What Makes JDs Fail

Failing JDSelling JD
"5+ years experience required""You've scaled teams from 10 to 50"
"Competitive salary""Top of market salary. We're profitable and hire the best"
"Fast-paced environment""You'll own outcomes for a $10M revenue line"
"Looking for a self-starter""You are a Marketer of Action"
"Great benefits""You'll work directly with [CEO name]"
Requirements listFuture vision of who they'll become

The Complete Asset Flow

Everything flows from the Job Scorecard:

StepWhat You CreateTemplate
Pillar 1Job ScorecardPillar 1
Asset 1JD Sales PageThis pillar
Asset 2Role VSLThis pillar
Asset 3Application FormThis pillar
Asset 4Recruiter PitchThis pillar

Derivatives: Channel-Specific Assets

Once you have the 4 core assets, you can create derivatives for each distribution channel your company has.

Example: LinkedIn Post

SectionWhat to Take
HeadlineThe hook question from Section 1
BodyIdentity statement from Section 2
CTALink to full JD

Your Unique Channels

Based on your company's unique distribution assets, you may create additional derivatives. Every company has channels most recruiters ignore:

If You Have...Use It To...
A communityPost roles where your biggest fans already hang out
A newsletterReach thousands who already trust your voice
A YouTube channelCreate role-specific content that lives forever
A product with usersIn-app announcements to people who love what you build
Founder social presencePersonal posts outperform company posts 10x

This is talent branding. Attract amazing people before you even post a job.

Below is the complete 10-section formula applied to a Wispr Flow ML Engineer role. Each section is labeled so you can see exactly how the template maps to the real JD.

1
The Hook

The Best Opportunity on Earth for an ML Engineer Who Wants to Kill the Keyboard

Are you ready to become the engineer who replaced typing for a billion people?

Imagine it's 2028.

You're known as one of the core engineers who built the voice layer that made keyboards obsolete. Your inference pipeline handles millions of requests at <500ms. Your models understand context so well that users forget they're talking to a computer.

You didn't just join another AI startup. You helped build the future of Human-Computer Symbiosis.

That's the opportunity at Wispr Flow.

2
Identity Statement

You're Not Just an ML Engineer

You've seen the hype. Everyone's "building AI." Most of it is wrappers on GPT. Throw more GPUs at the problem. Ship to cloud and call it a day.

That's not you.

You are a Voice Architect.

You get excited by constraints, not frustrated by them. You see "on-device, sub-500ms, zero edit rate" and think "now that's a real puzzle." You don't just train modelsβ€”you obsess over the 50ms that separates "magical" from "laggy."

You have the ability to:

  • Ship production ML systems that millions depend on
  • Debug across the entire stack (model β†’ API β†’ client)
  • Solve problems with engineering, not just by throwing compute at them
  • Make technical decisions that directly impact user experience
  • Move fast without breaking things that matter
3
Credibility Markers

Why Wispr?

You might have seen us on Product Hunt (#1 Product of the Day).

You might have tried Flow and thought "wow, this actually works."

The problem:

People type at 40 WPM but think at 400 WPM. The keyboard is a bottleneck to human potential. Voice interfaces have sucked for 30 years because they're slow, inaccurate, and awkward.

Our mission:

Replace the keyboard. Build voice interfaces so good that talking to your computer feels like talking to a close friend.

Where we are:

  • $81M total funding (latest $25M led by Notable Capital + Steven Bartlett's Flight Fund)
  • 50% month-over-month revenue growth
  • First voice dictation platform people use MORE than their keyboards
  • Team of 15 β†’ scaling to 50
  • Steven Bartlett partnership with The Diary of a CEO

Will you be the engineer who helps us get there?

4
Day-to-Day Specifics

What You'll Actually Do

This is not an "ML Engineer" role where you train models in notebooks and throw them over the wall. You will solve real constraint problems:

The Latency Problem

Sub-500ms from speech to text on screen. Audio β†’ spectrogram β†’ speech model β†’ LLM cleanup β†’ display. Every millisecond you shave, users FEEL it.

The Accuracy vs Speed Tradeoff

Bigger model = accurate but slow. Smaller = fast but worse. We need BOTH. You will get clever with distillation, quantization, architecture.

Context Awareness

Know if user is in Slack (casual) vs Email (formal). Not dumb transcription. Building something that UNDERSTANDS.

Personalization at Scale

Every user has different accent, vocabulary, style. Model adapts to each user over time. Learning from millions while respecting privacy.

On-Device Privacy

Users dictate sensitive content. Audio never leaves device. No cloud GPUs to lean on. Privacy is not a checkbox. It is a constraint that forces 10x better engineering.

The Last Mile Problem

95% β†’ 99% is easy. 99% β†’ 99.9% is brutal. Edge cases: rare accents, noise, jargon, mumbling. Anyone can build a demo. We are building something that works for everyone, everywhere, every time.

The 8 Speech Recognition Challenges You Will Solve

Most speech products solve 2-3 of these. We are going after all 8:

1
Short Audio Ambiguity: "Hey Lauren" vs "Hey Laurent" sound identical
2
Background Noise: Coffee shops, street noise, multiple speakers
3
Accent Adaptation: Boston, Mumbai, London - same word, different sounds
4
Domain Vocabulary: "Gabapentin 903" β†’ "gabapentin 900 3x daily"
5
Speech Understanding: Transcribe correctly AND understand meaning
6
Code-Switching: Users mix languages mid-sentence
7
Name Recognition: "Stephen Bartlett" requires knowing who that is
8
Incoherent Outputs: Catch outputs that are audio-plausible but wrong

You will work directly with Tanay (CEO), Sahaj (CTO), and a small, elite engineering team. You will make decisions that ship. No death by committee.

5
Success Picture (18 Months)

What Success Looks Like (18 Months)

  • βœ“You've cracked the latency problem - inference pipeline handles 10M+ daily requests at <200ms P95
  • βœ“You've solved accuracy vs speed - zero edit rate without sacrificing response time
  • βœ“You've shipped personalization at scale - returning users 40% more accurate, all on-device
  • βœ“You've pushed past the last mile - edge cases that broke competitors work flawlessly
  • βœ“You're known internally as "the person who made Flow fast"
  • βœ“You've helped grow the ML team from 3 to 10 engineers
  • βœ“Your code runs every time someone talks to their computer
6
Perks (Real Benefits)

What You'll Get

Compensation$130K-$240K + generous equity (we're early, it matters)
ImpactYour code will be used by millions of people every day
GrowthWork with exceptional ML engineers, not maintain legacy systems
OwnershipReal decisions, real autonomy, real accountability
MissionBuild something that actually changes how humans use computers
Team15 people who chose this over Google/Meta offers
7
Requirements (As Identity)

You'll Succeed Here If:

  • β€’You've shipped ML to production - not just notebooks, real systems with real users and real latency constraints
  • β€’You've optimized inference under constraints - you know the difference between P50 and P99 and why it matters
  • β€’You've worked on-device or edge ML - or you're excited to learn. No cloud to fall back on here
  • β€’You've solved the accuracy/speed tradeoff - distillation, quantization, pruning aren't just buzzwords to you
  • β€’You've worked at a startup - you know what "wearing many hats" actually means
  • β€’You obsess over the last mile - 99% β†’ 99.9% is where you come alive. Edge cases are puzzles, not annoyances
  • β€’You have strong opinions, loosely held - you'll debate architecture, then commit fully
  • β€’You want to build, not maintain - this is a building role

If you don't fit 100% but think you'd be great, apply anyway and tell us why.

8
Values (Real Filters)

Our Values (Real Ones)

High agency is on brand. Waiting for permission is off brand.
Obsessing over users is on brand. Obsessing over tech for tech's sake is off brand.
Sweating the details is on brand. "Good enough" is off brand.
Spirited debate is on brand. Passive agreement is off brand.
Shipping is on brand. Perfecting in isolation is off brand.
9
Clear Process

What Happens After You Apply

1
Application Review - We actually read these (24-48 hours)
2
Intro Call - 30 min with recruiting (tell us your story)
3
Technical Screen - 60 min with an engineer (real problems, not leetcode)
4
System Design - Deep dive on how you think about ML systems
5
Team Day - Meet the people you'd work with
6
Offer - Fast decisions, transparent comp
Skip the Line

Want to Skip the Interview Process?

We believe in proof of work over polish. If you want to jump straight to the front of the line, complete this challenge instead of the traditional application.

The Challenge (3-4 hours):

Your voice-to-text pipeline runs at 650ms end-to-end. Users complain it feels "laggy." Target: <500ms.

1. Identify where you'd focus optimization first and why

2. Propose 3 different approaches to get under 500ms

3. For each: What you'd do, expected savings, tradeoffs

4. Which would you try first? Why?

Submit your solution as a doc or GitHub repo. Strong submissions skip straight to a conversation with Tanay.

β†’This is how you put yourself in the top 1% of candidates.
10
Personal Close

A Note from Leadership

We're building something that matters. Voice is the most natural way humans communicate, and we're making it the most natural way to interact with computers.

If you're tired of incremental improvements at big companies, if you want your work to ship and matter, if you want to look back in 10 years and say "I helped build that" - we'd love to meet you.

- The Wispr Team

Pillar 2: Making Your Job Description Your Best Sales Page | Recruiting Playbook