Technical work simulations

A workday that reveals how someone works.

Candidates complete a realistic final assessment with role-specific tools and working context. Your hiring team receives the work, check results, and source-linked findings across software engineering, data & analytics, ML engineering, and ML research, then makes its own decision.

Try the 3-minute walkthrough View sample evidence Discuss a pilot

Inside the simulation

Four disciplines.
A real piece of the job.

Investigate an assignment, use the tools, and clarify requirements with AI coworkers. Explore an example from each role area.

ML engineering

Investigate a suspicious model improvement

Illustrative scenario

evaluate.py · illustrative excerpt

# Keep each customer in one split.
splitter = GroupShuffleSplit(
    n_splits=1, test_size=0.2, random_state=42
)
train, holdout = next(splitter.split(
    features, target, groups=customer_id
))
assert not set(customer_id[train]) & set(
    customer_id[holdout]
)
pipeline.fit(features[train], target[train])
prediction = pipeline.predict(features[holdout])

ML lead · AI coworker

The new score looks unusually strong. Can we trust the comparison with the baseline?

Candidate

I’ll check entity overlap, fit preprocessing only on training data, and compare both models on the same untouched holdout.

Example dialogue showing how working context enters the task.

See the actual simulation interface Software example · code & team chat

Actual SeeTheWork workspace showing Python webhook retry code beside a team conversation about compatibility and regression cases. — Actual interface · illustrative demo code and conversationThe task and enabled tools vary by role.

A closer look at the actual interface

Enlarged actual Code Studio showing Python retry logic. — Code Studio · scroll to inspect the full code

Enlarged actual Team Chat discussing compatibility, retry limits, and verification. — Team Chat · clarify and explain decisions

The process

A final assessment with a clear human handoff.

Choose a meaningful piece of the job, set consistent conditions, then review the evidence as a team. Follow-up can help when a question remains.

01
Shape the simulation
SeeTheWork drafts the task and review criteria. Your hiring manager supplies role context, checks realism, and agrees on the tools and conditions for a final assessment.
Starting point
A role brief and the work your final assessment should reveal
Review output
One approved assessment with consistent conditions
02
Candidate does the work
A real participant completes a roughly 90-minute assessment with realistic tasks, tools, and working context. The time limit and tool policy are agreed for the role before the session.
Starting point
An agreed task, participant scope, and session instructions
Review output
Finished work and the relevant working context
03
Make a human hiring decision
Your technical reviewer inspects the work and checks, challenges the summary, and identifies any useful follow-up. Your hiring team considers the evidence and makes the decision.
Starting point
Completed work and captured evidence
Review output
A review package for an accountable human decision

Two clear views

The candidate does the task. Your team reviews the work.

A short public walkthrough demonstrates the flow. A real pilot uses an agreed assessment session and its own candidate instructions.

Try the scripted 3-minute walkthrough

Candidate sees

Task and working context

Role-relevant instructions, available tools, and the task conditions agreed with your team. Real participant sessions are designed around 90 minutes.

Hiring team receives

Work with inspectable evidence

Candidate-produced artifacts, objective check results, source-linked findings, visible uncertainty, and possible follow-up questions.

The review package

A brief with sources you can inspect.

Reviewers can scan the summary or inspect the work behind it. Missing and uncertain evidence stays visible.

Explore all four evidence briefs

Finished work: The artifact or change the candidate produced.
Objective checks: Tests and task-specific results that can be verified directly.
Working context: Relevant decisions and evidence linked back to the work.
Human review: Reviewer notes, uncertainty, and the next questions to ask.

Start with one role and the work your final assessment should reveal.

Discuss a pilot

A workday that reveals how someone works.

Shape the simulation

Candidate does the work

Make a human hiring decision

The candidate does the task. Your team reviews the work.

Task and working context

Work with inspectable evidence

A brief with sources you can inspect.