Practical Assessments

Some positions include a practical assessment: a hands-on task you do directly in the browser, such as writing code, working in a spreadsheet, answering emails, drawing a system or process diagram, or drafting a document. This may be the only evaluation, or it may come after a conversational interview.

What to Expect

You'll see an overview page listing what's coming — how many problems and any time limits
Each problem shows a title and description explaining what to do
You work in a browser-based workbench that fits the task: a code editor, a SQL editor, a spreadsheet, an email inbox, a whiteboard for diagrams, or a text editor
You can drag the dividers between the instructions, your work and the results to make any of them bigger
Your work auto-saves every 30 seconds, so you won't lose progress
When you're done, click Submit Assessment

Coding Tasks

Pick any of the languages offered from the language menu. Your code in each language is kept separately, so you can switch back and forth.
Click Run to run your code against the problem's test cases. You'll see each test's result: passed, failed (with the expected and actual values), or an error. Running your code doesn't submit it.
Your code runs with no internet access. Only the standard library and a set of preinstalled libraries can be used.
If a test never finishes (for example, a loop that never ends), Run stops after a few seconds and tells you which test got stuck.
In Python and R, charts your code draws show in the console when you run it: call plt.show() with matplotlib, or draw with R's plots or ggplot2. Each chart is shown with its file name.

Web Page Tasks

Some coding tasks ask you to build a small web page in HTML and CSS with JavaScript or TypeScript.

Pick the language set you want from the language menu (for example HTML, CSS and TypeScript). Each set keeps its own files, so you can switch back and forth.
Your page appears live beside your code and updates a moment after you stop typing. Use the reload button above it to start it fresh.
Your files can import each other (import { products } from './products'). Packages and links to other websites aren't available: everything the page needs is in your files.
What your page logs with console.log, and any error it hits, shows in the Console below.
If a loop runs for more than a second, it's stopped with an error naming the file and line, so your page doesn't freeze. A page that stops responding altogether is stopped and tells you so; reload it after you fix the code.
There are no test cases: the reviewer reads your code and uses your page.

Robotics Tasks

Some assessments ask you to write a controller for a simulated robot, in Python or TypeScript.

Your control() function is called every control period with the robot's sensor readings and returns its commands. The read-only robot file lists the types and the helpers the task gives you.
Run all runs your controller on every scenario in your browser. Watch the run in the 3D view (play, or drag the scrubber), and read the Telemetry, Response, Results, Joints and Log tabs. What you print shows in the Log.
Sensors are noisy and slightly late, as on a real robot. Your controller is also graded on hidden variations of each scenario, so read goals and settings from the scenario rather than writing in the visible numbers.

SQL Tasks

The tables you work with are listed on the left. Each test has its own data: pick a test to browse its tables and click a table to see its rows.
Run query runs your query on the test data you picked and shows the rows it returns. It's for exploring and checking your work.
Run tests checks your query against every test and shows whether each one passed or failed, with the rows your query returned next to the rows expected.
The editor suggests table and column names as you type. Press Cmd/Ctrl+Enter to run your query.
Queries run on SQLite.

Diagram Tasks

You draw on a whiteboard: add boxes, ellipses and diamonds, type a label inside each, and connect them with arrows. Arrows stay attached to the shapes they connect.
Label your arrows with what flows over them, and add text notes to explain your decisions. Labels, arrows and notes are what gets evaluated; neatness and layout don't count.
Some tasks start with a few shapes already on the board for you to build on.

Assessments in Steps

Some assessments come in steps that build on each other, like real work does. The top of the screen shows which step you're on (for example Step 1 of 3).

A step can ask you to change what you just made (for example, "Finance now wants this by region"). That step opens on your own work from the step before.
A step can be a new task in the same workbench, or can bring your work into a new task, in another workbench (for example, emailing the spreadsheet you built) or the same one.
When your work is brought into a new task, it arrives inside that task's workbench, labelled with the step it came from: a spreadsheet's sheets become a CSV file in your mailbox's Files folder (open it, or click Attach to a new email), a file in your code project, a new sheet, a section of your document, or notes beside your diagram.
When you finish a step, click Hand In Step 1 of 3. Your work for that step is saved and you can't change it afterwards, and the next step opens right away in the same assessment.
Each step has its own clock, which starts when the step opens.
On later steps, the Earlier work tab beside the task shows what you handed in for the steps before, so you can refer back to it.
The last step's button is Submit Assessment. Your work across all the steps is reviewed together.

Talking With the Interviewer About Your Work

Some assessments include short conversations with the AI interviewer about your work, in text or by voice. The interviewer sits in a panel on the right, beside your task and your work, so you can see both while you talk. It looks and works like the interviewer in an interview. To make room, click the double arrow in its header to fold it away; a button in the corner brings it back, and the conversation keeps going meanwhile.

After each step: after you hand in a step, the interviewer asks quick questions about it while your next step is already open. The time you spend talking is added back to that step's clock. You may be able to skip these questions.
After the assessment: once everything is handed in, the interviewer asks questions about your work as a whole. Your work is final by then, and nothing you say changes it. When you're done, click End Conversation, then Finish Assessment.

Your AI Assistant

Some assessments give you an AI assistant you can chat with about your task while you work, as you would at work. Click Ask AI at the top of the screen (or press ⌘J, Ctrl+J on Windows): a panel opens out of the button, over your work. Drag it by its title to move it, pull any edge or corner to make it bigger or smaller, and fold it back into the button with the minus or Esc. It stays where you put it each time you open it again.

It reads the task and your work as it is right now, and, in an assessment in steps, what you handed in for the earlier steps and what you asked it before.
It can explain the task, check your work, find mistakes and write formulas, code, queries or drafts for you. Code and formulas come in blocks with a Copy button, and every message has a copy button under it.
Before you've asked anything, it shows a few questions you can send with one click. The message box grows as you write (Shift+Enter for a new line).
It only helps with this task: it won't answer questions about anything else.
You can send a set number of messages over the whole assessment; the button shows how many are left when you're running low.
The employer sees your conversation with it, and how you used it counts: check what it gives you before you use it.

Debugging Tasks

Some assessments show you something that is already broken (a spreadsheet, an email thread, code, a query, a diagram, a document or a robot's controller) and ask you to talk the interviewer through it. The Debugging strip above the work says it is Read only: nothing you do changes it, but you can click into every cell, line, email, shape and passage to investigate, and run code or queries where a Run button is offered.

The conversation starts as soon as the task opens, in the panel beside the work, in text or by voice.
Say what you see, what you think is wrong and why, how you would fix it, and how you would check that the fix worked. The interviewer won't tell you whether you are right.
To show the interviewer what you mean, select it in the work. The Debugging strip says what you can select in that workbench: for example a cell or a range, a line of code, an email, a shape, some words, or in a robot's run a stretch of a trace (pause the run and choose Select this moment to show the moment it is at). In text, your message box says what it is about (for example About B14 on Summary) and your next message goes with it; click its × to leave it off. By voice, tell the interviewer when you have selected something: it looks at what you last selected.
When the strip says how many problems there are, look for all of them.
When you're done, hand the step in as usual.

The interviewer won't do the work for you or tell you whether an answer is right. Answer as you would to a colleague: what you did, why, and what you would change. For a voice conversation your browser will ask to use your microphone; if the call drops, it starts again.

If There's a Time Limit

A countdown timer appears in the top bar
At 25% time remaining, you'll see a yellow warning
At 10% remaining, the warning turns red
When time runs out, your current work is automatically submitted — you won't lose anything

Tips

Read each problem description carefully before starting
If there's starter code or content provided, review it before making changes
Save manually anytime by clicking the Save button (in addition to auto-save)
If you have multiple problems, use the sidebar to navigate between them
Your camera remains active for environmental monitoring (same as the interview) — no video is recorded

Combined Assessments

If your position includes both an interview and a practical assessment:

You'll complete the conversational interview first
You'll see a transition screen — take a moment to prepare
Then you'll begin the practical assessment

The two parts are scored separately, then combined into a single evaluation.

Practical Assessments

On this page