Resumes describe.Code proves.
Candidates solve real problems in a real editor. Every submission is executed against the same weighted test cases, in the same sandbox, under the same clock — so you're comparing engineers, not adjectives.
No credit card required · Your first 5 assessment attempts are free
Given a list of request timestamps (seconds) and a window size w, return the largest number of requests that land inside any window of length w.
import java.util.*;
class Solution {
// Sliding window over sorted timestamps — O(n log n).
static int peakWindow(int[] times, int w) {
Arrays.sort(times);
int best = 0, left = 0;
for (int right = 0; right < times.length; right++) {
while (times[right] - times[left] >= w) left++;
best = Math.max(best, right - left + 1);
}
return best;
}
}Pedigree is a proxy.
Working code is the thing itself.
Every hiring process is a bet on a proxy — a school, a logo, a confident answer in a room. KodeQualify replaces the proxy with the artefact: the code they wrote, the cases it passed, the ones it didn't.
What a resume asserts
- Expert in six languages
- Five years at a company you've heard of
- Led performance optimisation initiatives
- CS degree, top-20 programme
Unverifiable. Ranked by whoever read it, in the order they happened to read it.
What a submission proves
- 8 of 10 weighted points, itemised per test case
- hidden #2 — empty input: expected 0, got 1
- O(n log n) — one million rows in 380ms
- 37 of 60 minutes used · plagiarism clean
Reproducible. Ranked against every candidate who answered the same question.
Skills over pedigree
Where someone studied doesn't compile. Rank on what the code did, then read the CV for context — not the other way round.
Real work, not trivia
Problems in a real editor against real runtimes, with the standard library available. If the job is Go, the test is Go.
One bar, every candidate
Same questions, same weights, same clock, same sandbox. Scores compare because the conditions are identical by construction.
Fair to the developer
Timed honestly, scored transparently, proctored — so the people who did the work aren't out-competed by the people who didn't.
From a question bank
to a decision you can defend
Author a question that can't lie
Write the problem and its test cases. Before it can be published, KodeQualify executes your reference solution in an isolated sandbox — a question that fails its own tests never reaches a candidate.
Send one identical bar
Bundle published questions into a timed assessment with per-question points. Every invitation carries the same set, the same weights and the same clock — no improvising per candidate.
- Peak request window4 pts
- Schema migration diff3 pts
- Rate limiter (multi-file)3 pts
Compare on evidence
Every submission runs case by case and scores itself. Watch the funnel fill, then rank on the numbers underneath — with the failing diff there when someone asks you to justify the call.
Every claim on this page
is something the platform does
// POST /api/code-test-questions → 202 Accepted
await createQuestion({
title: "Peak request window",
language: "go",
codeSnippet: referenceSolution,
testCases: [
{ input: "[1,2,3] 60", expectedOutput: "3",
isSample: true, weight: 1 },
{ input: "[] 60", expectedOutput: "0", weight: 2 },
],
});
// pending_validation → validating → publishedQuestions are executed before anyone is invited
Author the problem, the reference solution and the weighted cases together. KodeQualify runs them in a sandbox and only publishes when the solution passes its own tests — so a broken question can never cost a candidate the role.
{
"candidate": "marco@example.com",
"assessment": "Backend Engineer · Screen",
"totalScore": 8,
"maxScore": 10,
"results": [
{ "case": "hidden #2", "passed": false, "weight": 2,
"expected": "0", "received": "1" }
],
"elapsedMinutes": 37,
"plagiarismSignal": null
}Scored per test case, not per gut feeling
Partial credit is the point. Each case carries its own weight, so a candidate who solves the problem but misses one edge case reads as 8/10 — with the failing diff attached — instead of a shrug and a maybe.
# identical conditions, every candidate, every attempt
runtime go1.23 # pinned, not "whatever's installed"
cpu 2 vCPU
memory 2048 MB
network deny # no egress, no lookups
timeout 5000 ms
filesystem fresh microVM, destroyed on exit
# a faster laptop earns no points hereThe same sandbox grades everyone
Every submission runs in a single-use microVM with a pinned runtime, fixed CPU and memory, no network and a hard timeout. Nobody is scored on their laptop, their connection or the hour of day they sat the test.
Choose a plan
that works for you
From interns to senior engineers across every tech role, KodeQualify helps you cut time-to-hire by 50% or more and hire with confidence.
Starter
The basics to start hiring
billed monthly
- 1 user
- Access to Screen + Interview
- Advanced plagiarism detection
- Leaked question protection
- Multi-file project questions
- 2,000+ questions
5 candidates included
then $20 per candidate invited
Pro
Everything in Starter, plus
billed monthly
- Unlimited users
- Advanced AI features
- AI-Assisted IDE
- AI proctoring & identity verification
- Advanced evaluation & Scorecard Assist
- Greenhouse, Lever & Ashby (ATS)
- Google & Outlook calendar sync
- 4,000+ questions
25 candidates included
then $20 per candidate invited
Hiring needs vary by month? You only pay for the candidates you actually invite — no seat counts to forecast, no unused allowance to write off.
Teams that hire on evidence
From six-person startups to campus drives with a hundred thousand candidates — anywhere the decision has to hold up after it's made.
We screened 1,100 applicants for six backend roles in three weeks. The shortlist wasn't a debate — it was a sorted column, and every engineer on the panel could see exactly why each name sat where it did.
Sarah Chen
Head of Talent, Nimbus
We stopped arguing about whether someone 'seemed strong.' The failing case is right there in the report. First-year attrition halved, because we started hiring people who could do the work rather than describe it.
Marcus Rodriguez
VP Engineering, GradeLab
Candidates tell us it's the first screen that felt like the actual job — same editor, same runtime, real problems. And they get a score back instead of silence, which is the part they thank us for.
Emily Watson
Engineering Manager, Plugable
Your next hire is already
writing the answer
Build one assessment, invite five candidates, and read the difference between what a resume claims and what a submission proves. Your first five attempts are free.
- Weighted test cases
- Plagiarism detection on every attempt
- ATS-ready exports