Skip to content
buildinterview
Sign inPricing

Work-sample interviews for data & AI teams

See the work.
Know the
candidate.

Give data and AI candidates work that reflects the job. See their thinking, AI-assisted execution, and the evidence behind their decisions.

Explore an Audition Talk through your role
Step inside the candidate experience
AI-generated illustration of a professional thoughtfully working at a laptop.
AI-generated illustration
Audition illustration
Data AnalystAI off

First, their own thinking.

Frame the problem Make assumptions explicit Decide what to verify
The work and the reasoning stay connected.

A different role. A different kind of work.

Explore role coverage

Data AnalystDiagnose a business change, reconcile the metric contract, and recommend the next decision.

The candidate experienceInteractive product walkthrough
Illustrative demo
ExitCheckout margin incident
Question 1 / 3 46:12 Pause
Plan Build Writeup
34 min elapsed
  1. import pandas as pd
  3. sessions = pd.read_csv("data/sessions.csv",
  4. parse_dates=["started_at"])
  5. orders = pd.read_csv("data/orders.csv",
  6. parse_dates=["created_at"])
  8. # inspect the change before writing findings.json
  10. print(f"Loaded {len(sessions)} sessions / {len(orders)} orders")
main Illustrative workspace0.4%PYanalysis.py local
Explore the files, review an AI edit, or run the demo.Example content · simulated AI and execution

Built around the work.
From the first thought to the final decision.

Independent
Plan
AI-assisted
Build
Decision-ready
Writeup
Evidence-linked
Review

The Audition

Give candidates the job.
Review how they do it.

Three connected phases reveal how someone frames a problem, works with AI, and turns the result into a decision.

  1. 01AI off

    Plan

    Their thinking, first.

    Candidates frame the problem, state their assumptions, and decide what to check. Their submitted Plan stays visible throughout the work.

    Independent judgment
  2. 02AI on

    Build

    The tools to do the job.

    A working project with files, data, code, a terminal, and a workspace-aware assistant. Candidates review changes and verify their work.

    Execution and verification
  3. 03AI on

    Writeup

    A decision they can explain.

    The final submission connects the recommendation to the analysis, limitations, and next steps. The work and the reasoning stay together.

    Ownership and communication

One interview version. One connected record of the work.

See the evidence

Built for your team

Different roles.
Different kinds of great work.

Design around the actual job. Explore eight role families, from Data Analyst to AI Engineer, with seniority and business context kept distinct.

Coverage modelShape the interview around the job
Illustrative configuration
Seniority
Selected role contractSenior · Ecommerce

Senior Data Analyst

Diagnose a business change, reconcile the metric contract, and recommend the next decision.

Expected scopeAmbiguous decisions with business consequence
Evidence to collect
  • Metric judgment
  • SQL reasoning
  • Decision communication
Possible composition
AuditionSQLMCQ

Availability, validation state, accessibility constraints, and content version remain visible before selection.

Audition

The complete work sample

One business problem. Plan, Build, and Writeup.

Python

Focused assessment

Analysis and engineering with executable outputs.

SQL

Focused assessment

Queries, joins, and reasoning about data.

MCQ

Focused assessment

Role-relevant foundations and judgment.

The reasoning behind the result

A finding is only as useful
as the work behind it.

Move from the evaluation to the Plan, code, execution, and final submission that support it. Keep human review grounded in inspectable work.

Candidate result · illustrativeWhy this finding was produced
Evidence linked
Evaluation finding

The candidate challenged the measurement contract before recommending a rollback.

The conclusion is tied to inspectable work. Activity volume alone does not establish the finding.

4 source types connected
Plan.md · submitted 09:14Source 1 of 4
01Confirm the checkout event contract before treating the alert as a product effect.
02Separate traffic mix from within-segment change.
03Reconcile orders, payments, discounts, and refunds before recommending action.
Preserved before the workspace assistant became available.

Select a source to follow the finding back to the work. Illustrative example.

For the hiring team

One role.
One consistent
interview.

Set the context, choose the assessment, and lock the interview version. Invite candidates under the same conditions, then review their work.

  • Role-specific interview design
  • Defined AI and timing policies
  • Evaluation with supporting evidence
Open company sign in
Role workspaceIllustrative configuration
Interview design Version 01

Senior Data Analyst

Data & analytics / Ecommerce

Role brief2 Interview design3 Candidate delivery
AuditionPlan, Build, and Writeup
Work sample
SQLFocused data reasoning
Focused
Candidate AI
Company-configured
Timing
One whole-test limit
Evidence
Complete profile
Review
Evidence-linked
Conditions stay bound to the interview version.

Optional product deep dive

Real context.
Connected work.

Explore the fictional Ecommerce scenario behind the demo. Follow its data model into SQL, Python, and a final decision memo.

Open the complete walkthrough
ordersOrder-level context
analysis.pySQL & Python work
Decision memoEvidence & next steps
Explore the demo schema9 tables · 2 Build tasks
Checkout margin AuditionPurpose-built public walkthrough · not bank inventory
Tables
9
Relationships
12
Build tasks
2
Demo contract mapped
01 · Inspect the demo data

The schema gives the work its context.

Active question path
Selected tableorders1.8m rows3 visible fieldsUsed by the selected work sample
02 · Trace the Build work

SQL and Python use the same world.

SQL40 min suggested workload

Business decision

Did the mobile release reduce true net revenue, or did reporting definitions create the alert?

Bound data

customerssessionsorderspaymentsrefunds

Candidate deliverable

A grain-safe reconciliation with explicit revenue, refund, and device definitions

The tasks belong to one Audition. The candidate must reconcile the same business decision across SQL, Python, and the final memo. No private checks or answers are exposed.

Switch tasks to trace the exact tables used by each part of Build.Demo-only content · no qid · no private question or answer material

Clear session rules

Evidence requirements are disclosed. Partial or failed evidence stays visible in review.

Versioned conditions

Content, runtime, AI, timing, and policy stay tied to the interview version.

Protected assessment content

Private graders and answer material are kept out of the candidate workspace.

AI-generated illustration of two technical colleagues discussing their work.
AI-generated illustration

The next great person on your team.

Make the conversation
about the work.

Step inside an Audition. Then build an interview around your role.

Explore the demo Talk through your role