For AI data company roles

What is a Strategic Project Lead (SPL)?

An SPL runs an AI lab's data project end to end: the plan, the experts, the quality and the budget. The interview looks a lot like a consulting case, and you can practice it here.

Start the SPL case

Free trial included.

Built with former SPLs from Mercor, Scale AI, Handshake, Turing and more

model.xlsxSummary
Your numbers
Tasks acceptedC56,000
Cost per accepted taskC8$512
Price per accepted taskC9$695
Gross marginC1026.3%
Experts needed, week 7C14118
M

Marcus · Head of Operations, Ashgrove Data

Walk me through your $512 per accepted task. Where do rework and discards sit in that number?

Hold to talk13:42 leftNo hints

What SPLs do

A Strategic Project Lead runs data projects for AI labs at an AI data company. The lab says what training data it needs. The SPL turns that into a plan, finds and manages the experts who make the data, keeps quality high, and delivers on time and on budget. Many postings ask for consulting or finance experience, and the interview looks a lot like a consulting case.

  • Own the client project.

    Be the AI lab's main contact and run its project from the first spec to final delivery.

  • Design the data pipeline.

    Turn what the lab asks for into tasks, guidelines and a weekly delivery plan.

  • Find and manage the experts.

    Source, vet, train and manage the contractors and domain experts who make the data.

  • Hold quality and margin.

    Set up quality checks, price the work, and keep the project on budget.

Companies hiring SPLs

Posted pay and what candidates report about each interview. Every fact says where it comes from.

Mercor

Strategic Project Lead (also written Strategic Projects Lead), and Strategic Project Associate for new grads

Pay: An SPL posting listed a $120K to $200K base. An associate posting listed a $120K base. Both add equity.

Interview: Candidates on Glassdoor report a case and dataset sent about 24 hours ahead, a live round on your proposal, an onsite in San Francisco and a final round with operations leadership.

Mercor SPL interview →Mercor SPA interview →

Scale AI

Strategic Projects Lead

Pay: A Generative AI posting listed $121,600 to $190,000 a year plus equity, for San Francisco, New York and Seattle.

Interview: Candidates on Glassdoor report a phone screen, a consulting-style case, a simple SQL round and behavioral interviews.

Scale AI SPL interview →

Handshake AI

Strategic Projects Lead, and Strategic Projects Associate in India

Pay: A San Francisco SPL posting from May 2026 listed $160K to $185K plus equity.

Interview: The India associate posting describes two steps: a short interview, then a day working on a case and presenting it, in the office or over Zoom.

Surge AI

Strategic Projects Lead

The posting describes running data operations end to end and setting data quality metrics.

Interview: We found no public interview reports.

Turing

Strategic Project Lead, Software Engineering (New York, San Francisco, Seattle) and Strategic Project Lead, Code (India, remote)

As listed on Turing's careers board in November 2026.

micro1

Strategic Project Lead

The posting describes the main contact for AI lab customers, managing a team of domain experts. Remote.

Interview: micro1 says every candidate goes through an AI interview with its AI recruiter, Zara.

Labelbox

Strategic Project Lead

The posting describes owning day-to-day delivery of customer data projects, including margins.

Pay: A California hybrid posting listed $140,000 to $170,000.

Titles, pay and openings change, so check each company's careers page. CaseTutor is not affiliated with any of these companies.

The SPL case interview

Candidates for Mercor's Strategic Project Lead role report a written proposal case: a client brief and data, then a memo and a spreadsheet model in about 24 hours, then a live round where interviewers push back on it. Some also report a second case they have not seen before. Other AI data companies run a spoken, consulting-style case instead; see each company above. CaseTutor gives you an original case in the written shape, an AI interviewer who attacks your numbers, and a graded report. In the app it is called Proposal Defense.

  1. Step 1

    Get a client brief and data.

    Pick one of two original cases. You get the client's brief, the data and a spreadsheet template.

  2. Step 2

    Write the memo and build the model in about 24 hours.

    Write the memo in the editor, or upload a .docx or .pdf. Build the model in the .xlsx template, in Excel or Google Sheets. Plan for 4 to 6 hours of work.

  3. Step 3

    Defend it to an interviewer who attacks your numbers.

    The interviewer is AI. It questions your own numbers and assumptions, live, and it gives no hints.

What's graded

Two graders, one report. Scores stay hidden until the defense is over.

The memo6 dimensions

  • Data type
  • Expert sourcing and vetting
  • Quality control
  • Ramp plan
  • Billing and margin
  • Writing

The defense4 dimensions

  • Holding ground with evidence
  • Conceding well
  • Numbers under pressure
  • Operational realism

Scores stay hidden until the defense is over. Then one report shows:

  • your memo scores
  • spreadsheet issues, with cell references
  • moments from your defense, with quotes
  • one thing to fix

Your report

Submitted after 21 h

  • Data type4/5
  • Sourcing and vetting3/5
  • Quality control4/5
  • Ramp plan2/5
  • Billing and margin4/5
  • Writing5/5

Summary!C14 · Experts at week 7 do not cover the week 7 task volume.

Held with evidenceDefense, question 1

You: “It is 2.4 author hours and 0.6 review hours per accepted task, at the pilot pay rates on the data tab.”

One thing to fix: your ramp hires for week 1 volume. Size the team for week 7 and work back to the start date.

Interview this week? Open the brief today and defend it tomorrow.

Free trial included.

Start the SPL case

The two cases

Pick either one. Both follow the same format and get the same report.

A mid-size model labOriginal case

Quillmoor Labs

Training tasks for a spreadsheet agent

Scope, staff and price 6,000 expert-written tasks in 10 weeks, inside a fixed budget.

A voice AI startupOriginal case

Saltmarsh AI

Preference data for a six-language voice agent

Price and staff 180,000 rated comparisons in six languages in 8 weeks, with one language short of raters.

Both clients are invented. In each case you play a Strategic Project Lead at Ashgrove Data, a fictional data vendor.

Our rules

They match the real format, so the practice counts.

  • No AI writing help.

    Write the memo and build the model yourself.

  • No hints during the defense.

    The interviewer pushes back and does not coach you.

  • Scores stay hidden until the end.

    You defend your proposal without knowing how the memo scored.

  • About 24 hours, like the real format.

    Late work is still accepted. Your feedback notes the time you took.

Frequently asked questions

Are these real company cases?

No. Both cases are original. They follow the shape candidates describe: a brief, a memo, a model and a live defense. The clients, the data and the numbers are invented.

How long does it take?

About 24 hours is suggested for the memo and the model, as in the real format. Plan for 4 to 6 hours of work. Late work is accepted, and your feedback notes the time taken. The live defense comes after you submit.

Is the interviewer AI?

Yes. The interviewer is AI, not a person. It questions your numbers and assumptions, and it gives no hints.

Which spreadsheet tools work?

Excel or Google Sheets. You build the model in the .xlsx template we provide and hand it in as .xlsx. In Google Sheets, use File > Download > Microsoft Excel (.xlsx).

How do I hand in the memo?

Write it in the editor in the app, or upload it as a .docx or .pdf.

When do I see my scores?

After the defense. Scores stay hidden until then. Your report covers your memo scores, spreadsheet issues with cell references, moments from your defense with quotes, and one thing to fix. If you skip the defense, you still get the memo and spreadsheet report.

Who can use it?

CaseTutor members, including the free trial. Each defense uses one interview credit.

Practice the whole loop before the real one

Open a brief, write the memo, build the model, then defend it. Open to CaseTutor members, free trial included.

Start the SPL case

Former SPLs from Mercor, Scale AI, Handshake, Turing and other AI data companies helped build these cases.

CaseTutor is an independent practice tool. It is not affiliated with, endorsed by, or sponsored by Mercor, Scale AI, Handshake, Surge AI, Turing, micro1 or Labelbox. The cases are original and do not reproduce any company's interview material.