Coding assessments

Watch candidates write working code

Real execution against your test cases in 13 languages, inside a sandbox with enforced limits. Not multiple-choice questions about code.

Start free See pricing

Run-before-submit
Hidden test cases
Cohort plagiarism detection

How it works

Execution, not estimation

A coding question is a problem statement, a set of test cases and resource limits. The score comes from running the code.

An editor candidates recognise

The same editor engine as VS Code, self-hosted so nothing loads from a third party mid-assessment. Syntax highlighting, autocomplete, and a layout the candidate can arrange.

Your test cases decide the score

Write visible cases so candidates can check their work, and hidden cases so they cannot reverse-engineer the expected output. Score is the proportion of cases passed.

Limits the sandbox enforces

Per-question CPU time and memory ceilings, applied by the container rather than by the browser — so an infinite loop fails the case instead of hanging the attempt.

Nothing is lost to a dropped connection

Code is saved as it is typed. A candidate who loses power resumes on the same attempt with their work intact, and it still counts as one candidate.

Plagiarism across the cohort

Submissions are compared to each other by token sequence, syntax tree and structural shape, with matching regions shown side by side for you to judge.

Isolated from everything

Execution happens in a container with no network access, separate from the application. Candidate code cannot reach your data or ours.

Languages

Published runtimes, not a vague count

The exact interpreter or compiler each language runs on, because a candidate's code either works on that version or it does not.

Compiled and interpreted languages side by side in one assessment
Per-question language restriction, so a Go role tests Go
SQL executes against SQLite and is checked on query output

Runtimes

C (GCC 9.2.0)
C++ (GCC 9.2.0)
Python (3.8.1)
Java (OpenJDK 13)
JavaScript (Node 12)
TypeScript (3.7.4)
Go (1.13.5)
Rust (1.40.0)
Kotlin (1.3.70)
Swift (5.2.3)
PHP (7.4.1)
Ruby (2.7.0)
SQL (SQLite 3.27.2)

Integrity

Similarity you can inspect, not a number to trust

Plagiarism detection compares every submission in the cohort against every other one, using three independent methods so a renamed variable does not defeat it.

Token sequence comparison catches copied code with identifiers changed
Syntax tree comparison catches reordered but structurally identical solutions
Structural comparison catches shared shape across different surface syntax
Matching regions are highlighted side by side for a human to make the call

Token sequence

Resists renaming

Syntax tree

Resists reordering

Structure

Resists rewriting

A match is evidence for a person to weigh, never an automatic disqualification.

Questions

Before you build your first problem

Can candidates run their code before submitting?

Yes. They can run against the visible test cases as often as they like, which is how a real engineer works. Hidden cases are only evaluated on submission.

How are partial solutions scored?

By the proportion of test cases passed, so a candidate who handles the main path but misses an edge case scores accordingly rather than getting zero. You control the weight of each question.

Do you support SQL questions?

SQL is one of the execution languages, running against SQLite. You write a problem and test cases, and the candidate's query output is compared to the expected output. There is no visual schema designer or result-set diff view — if you need those specifically, it is worth knowing before you commit.

What stops candidates pasting in an answer from elsewhere?

Copy, cut and paste inside the assessment are recorded as integrity events and feed the attempt's risk score, and plagiarism detection compares submissions across the whole cohort. Neither is a claim to prevent it outright — they make it visible so you can decide.

Which language versions run?

The exact runtimes are listed on this page. We publish versions because a candidate writing Python needs to know which interpreter their code will hit.

Set a coding round live this week

Write a problem, add test cases, invite a candidate. The free trial covers a real drive.

Start free trial View pricing & plans

Have questions? Email sales@parikshafy.com