The AI drafts. You decide.
First-round technical screening for intern and graduate software roles. AI prepares the questions and a draft rating for each skill. Nothing reaches a candidate until you approve it, and every rating shows the code or answer it came from.
Evidence report · Pushkar Kirange
Backend Intern · Round 2, codingCriterion 2 of 5: Handles edge casesSample dataEvidence
if len(a) != len(b): return False return Counter(a) == Counter(b)
“I’d compare the lengths first. If they differ it can’t be an anagram, and that check is constant time.”
9 of 10 passedTest 10 failed: the same accented word in two Unicode forms.
Rating
AI draftRejects length mismatches early E1 and explains why E2. Did not normalise Unicode input E3.
Drafted from rubric v3. Not a decision.
Decision
HumanTest 10 is outside the stated constraints (lowercase English only). Not counting it against her.
Priya Nair, Recruiter · Nov 3, 14:20
Eight steps from job post to decision. You sign the ones that matter.
AI does the drafting. Candidates do the work. Between them sits your approval, and at the end sits your decision.
- 1, done by you
Create the job
Role, skills to test, time limits.
- 2, AI draft
AI drafts rounds
Questions, rubric, test cases.
- 3, done by you
You approve
Edit anything, then publish a fixed version.
- 4, candidate
Invite
Consent, device check, a practice question.
- 5, candidate
Interview
Text or recorded voice, with the camera off.
- 6, candidate
Coding
Real editor, sample tests, autosave.
- 7, AI draft
Evidence report
Test results plus draft ratings that cite evidence.
- 8, done by you
You decide
Confirm or change each rating, then advance or decline.
Nothing AI-drafted crosses the gate without you. Approved versions are locked while candidates are taking them, so every candidate in a cohort answers the same questions.
For candidates, it should feel like a fair test, not surveillance.
Most of your applicants are students taking their first technical screen. A calm, predictable room gets you their real work, and more of them finish.
- Camera stays off
- Interviews are text or recorded voice. We say exactly what's recorded and why before anything starts.
- Practice first
- A device check and a practice question before any timer runs.
- Nothing lost, nothing hidden
- Visible autosave, a server-kept timer that never resets, and the hidden-test policy stated up front.
- Built to be accessible
- Keyboard and screen-reader friendly, captions on interview questions, and accommodations requested right from the invite.
Hi Chetan, one round left
- Invite
- Interview
- Coding
- Decision
Two short Python problems. Start whenever you're ready before Nov 14.
Send link to my laptopWhat we refuse to build.
Some hiring AI is sold on what it can read from a face or a voice. We think that's the wrong product, so these limits are built into HireArena as rules that no account setting can switch off.
No face, emotion, eye-movement or accent scoring.
How someone looks or sounds is not a skill. We don't measure it.
No guessing protected traits.
Not age, gender, ethnicity, disability or anything like them, from any signal.
No automated rejections or offers.
Every advance and every rejection is made by a person, and their name is on it.
No AI question a candidate sees before you approve it.
Drafted questions stay in your review queue until you publish them.
No single opaque score.
Ratings are per skill, and each one points to the code or transcript behind it.
Run a pilot cohort with us.
We're looking for a few design partners: campus recruiting teams and small engineering teams hiring interns, graduates and trainees.
- One junior software role, about 100 candidates.
- Two rounds: a short interview and a coding problem.
- You approve every question and make every decision.
- Candidates never pay to take your assessment.
HireArena is in pilot. You won't find customer logos or success numbers on this page yet. The pilot is how we earn them, and your feedback shapes the evidence report.