Correctly Defining the Role

Core Philosophy: We're looking for a person that has the highest likelihood of achieving an outcome.

If we can define this persons outcome to 2 variables, this will give us a high likelihood of confidence when we interview them.

Variable 1: The Non-Negotiable

The single most important capability. If they have this, you can teach the rest.

This person has [VERB] + [OUTCOME] + [CONSTRAINT]

Questions to figure out Variable 1:

1. What's the #1 outcome this person needs to achieve in 6-12 months?
2. What past experience proves they can do this?
3. What makes this hard here specifically? (the constraint)
4. What interview question would verify they've done this?

Variable 2: Team Building (if applicable)

Can they attract and develop talent for their function?

This person can attract [ROLE] who [QUALITY]

Note: Defining a person within two variables does not mean this is a binary process. It does not mean that if someone does not have these exact variables that we would not hire them. Every conversation that you have to bring someone incredible onto your team is more about "what is the best seat for this person given that they are great?"

The Knockout Question

Ask this in every interview to validate Variable 1:

"Tell me about a time you [DID VARIABLE 1]. What were the constraints? What was the outcome?"

If Variable 2 is in play: "Tell me about someone you hired. How did you find them? What made them great?"

Remember: It is OK if they do not meet 100%. We are always asking: "What is the best seat for this human?"

Why Assess the Team?

You already have great people. Study them to understand what "great" looks like. Then hire to fill the gap, not duplicate what you have.

Step 1: Interview Current Team Members

For each person in or near the role, ask:

1. What makes them great? (not just skills - their personality superpower)
2. What's their day-to-day? (the new hire needs to be great at this)
3. What problems are they actually solving?
4. Why is this their dream role? (or what would they like to see improve to make it their dream role)

Bonus: Have the team record their answers so you can later use these as selling assets to send to candidates before interviews.

Step 2: Find the Gaps

This is about understanding where the team is strong and where there are gaps this person can fill.

Gaps can be:

1. Technical - Missing skills, domain expertise, or experience with specific tools
2. Vibe/Energy - Maybe the team is super serious and you need someone who can brighten the room
3. Working style - Fast shippers vs deep thinkers, async vs sync communicators

Questions to ask:

1. What skills does one team member have that others don't?
2. What's common across all of them?
3. What's missing that would complement the team?
4. What personality trait would balance the current dynamic?

Important: This is an assessment, not a sole hiring criteria. You don't hire someone just because they fill a gap. But understanding the gaps helps you know what to look for.

Step 3: Compare Team Members

If you have multiple people in similar roles, pin them against each other:

Questions that actually reveal gaps:

1. What problems does the team struggle with that nobody is great at solving?
2. When the team gets stuck, what type of person would unblock them?
3. What's the thing everyone avoids because nobody owns it?
4. What expertise would make the team 2x faster?
5. What personality trait is missing from the current dynamic?

The Talent Density Rule

Every hire should raise the bar. At minimum, they should be just as good as the people on your current team. But you want to strive to hire people that are better. Over time, you're hiring better and better people. That's how you build a world class team.

Completed Scorecard: ML Engineer

Role Overview:

Vision for the Role:

Build the ML system that powers millions of users writing faster than ever. You'll work on on-device speech recognition, pushing the limits of latency and accuracy. This is the core technology that makes Wispr Flow magical.

Single Sentence Filter:

This person has shipped on-device speech recognition with sub-500ms latency to production users.

Candidate Profile:

1. On-device shipping - Has deployed models to constrained devices (phones, not cloud). Knows the difference between notebook accuracy and production accuracy.

2. Latency obsession - Understands sub-500ms is the bar. Thinks in P50 vs P99. Has debugged why some users see 1200ms while others see 400ms.

3. Accuracy/Speed navigation - Has made the hard tradeoff. Knows when to sacrifice 2% accuracy for 100ms. Understands distillation, quantization, pruning.

4. User outcome focus - Measures success by zero edit rate, not WER benchmarks. Knows "good WER" and "users don't have to edit" are different things.

5. Privacy-first mindset - Defaults to on-device. Understands why cloud is not an option. Has worked within memory and compute constraints.

6. Edge case hunter - Obsesses over the last mile (99% → 99.9%). Knows the 8 accuracy killers: accents, background noise, code-switching, domain vocabulary, short utterances, name recognition, speech understanding, incoherent outputs.

Dream Role Overview:

1. You own the full stack (on-device, not cloud). No waiting on other teams.
2. Clear success metric: zero edit rate. You know exactly when you've won.
3. Sub-500ms latency is the bar. Real engineering constraints, not vanity benchmarks.
4. Your improvements ship to millions of users. You'll feel the impact in days, not quarters.
5. Work directly with founders. No layers of management between you and decisions.

Required Skills & Responsibilities:

Knockout Questions:

Business Goal:

Build the ML system that powers millions of users writing faster than ever. Zero edit rate is the north star.

This completed scorecard feeds into: JD, outreach messaging, interview questions, and discovery project criteria.