The product

A work simulation. Evidence for your team.

Four stages, from authoring the challenge to the decision. All of it happens in the browser: your team installs nothing, and neither does the candidate.

Adaliza mascot: a robot taking notes on its clipboard with a pencil

The challenge is authored

A tech lead describes the opening, the stack and competencies to evaluate. An assistant proposes a draft and refines it. Approval is always human.

The scenario and the role the agent plays.
Which information stays hidden, and the condition that releases it.
The rubric: competencies, observable criteria and weights.
Inviting fixes the case and rubric to one version for that group of candidates.

Configuration of the challenge "Multi-branch sales pipeline with Medallion architecture": statement, context, deliverables, agent role and stack. — A real Adaliza screen (Spanish interface), with example data.

The candidate is invited

The invitation is individual: their own link, with an expiry date and a single use. No account, no password.

Name and email are the identifying fields requested.
The link stops working if it expires, is cancelled, or the challenge is already complete.
The team sees the state of every invitation in the console.

Tracking for the Data Engineer vacancy: six candidates with their status, dates and soft and technical scores. — A real Adaliza screen (Spanish interface), with example data.

The candidate works through the simulation

They open their link, read the case, and talk to the agent. It is deliberately incomplete: the agent holds what is missing.

Before starting they are told they are talking to an AI agent and being evaluated.
The agent releases hidden information only when the condition the team configured is met.
They see the remaining time and the submission conditions from the first screen.
They submit their solution as rich text or a repository link.

The candidate's conversation with the agent playing the head of Commercial Analytics. — A real Adaliza screen (Spanish interface), with example data.

Results and evidence

The conversation and the deliverable are recorded. The evaluation walks them against the rubric. You see a score with the quotes behind it.

Technical and soft skill scores, plus scores per criterion, on a 1 to 5 scale.
Every criterion with the fragments of the conversation or the deliverable that support it.
The automatic score is read-only; human observations go in separate fields.
Comparison across candidates holds within the same challenge and the same configuration version.

The candidate's result: soft and technical skill scores, summary and score per criterion. — A real Adaliza screen (Spanish interface), with example data.

Technical skills evidence: each score with its justification and the quote from the deliverable. — A real Adaliza screen (Spanish interface), with example data.

What you can review

The solution and the decisions that shaped it.

The deliverable

The solution they submitted for the problem.
Technical decisions documented in text or in the repository.
How the work addresses the rubric’s criteria.
Submission excerpts that support the evaluation.

The recorded process

What context they requested before proposing a solution.
Which assumptions they stated and which alternatives they discussed.
How they explained their judgment to the simulated counterpart.
Conversation excerpts that support each score.

Every candidate, on the same rubric.

The comparison holds within the same challenge and version. Adaliza marks where candidates differ most.

Candidate comparison by criterion, with the column where they differ most highlighted. — A real Adaliza screen (Spanish interface), with example data.

The Adaliza mascot: a robot waving, holding a clipboard

Let us walk it through on one of your openings.

We build the challenge on a real role and go through it together, stage by stage. Thirty minutes is enough to know what it measures and what it does not.

Book a demo