NorthAssay
Product

Product

  • Verified answersThe interview asks about what they submitted.
  • How it worksFrom a job post to the evidence, in three steps.
  • Judgment evaluationMeasure the calls the job actually leaves open.
  • AI video interviewsLive conversation, not a one-way recording.
Pricing
Guides

Guides

  • Assessment methods

    • Candidate assessment
    • Skills assessments
    • Technical skills assessment
    • Communication assessment
    • Pre-employment assessment
  • Tools & software

    • Assessment tools compared
    • AI candidate screening
    • Hiring software for small teams
  • By role

    • Technical interview questions
    • Assess a DevOps engineer
    • Assess a data engineer
    • Assess an engineering manager
  • Screening practice

    • Five checks
    • When candidates use AI

Trust

  • Responsible AIWhat the score is, what it is not, and who decides.
  • SecurityWhere candidate data lives and how long it stays.
  • AboutWhy NorthAssay exists.
Sign inStart free

Anyone can do the task. Few make the call.

Paste the job. NorthAssay interviews candidates live on the judgment calls it leaves open, following up on what they actually say. Know who’s right for a freelance or remote job you’ll never watch them do.

Paste a job description
northassay · scoreboard

Demo: NorthAssay reads a role brief, reasons through it step by step, drafts a tailored assessment you can review and save, then scores candidates with rationale you can read, requirement by requirement.

  1. Brief

    Senior Backend Engineer · Go + Postgres · leads a small team

  2. Senior Backend Engineer · Senior · 5 competencies

    Senior backend role. Go and Postgres, leads a small team. Weighting system design and ownership over syntax trivia.

    GoPostgres / SQLSystem DesignTeam LeadershipObservability
  3. 3 assessments planned

    Three instruments: MCQ for fundamentals, a written prompt for judgment, a video interview for communication.

    Core fundamentals

    MCQ · Senior · 8q

    System design judgment

    WRITTEN · Senior · 3q

    Communication

    VIDEO_INTERVIEW · Senior · 4q

  4. 3 questions for System design judgment

    Drafting a written prompt tuned to this role, not pulled from a template library.

    Written prompt

    Scale a write-heavy service to 10× traffic. Walk through your data model, indexing, and the failure modes you’d watch.

  5. Ready for review

    Assembling the agent, the three assessments, and every drafted question into one reviewable draft.

    Assessment· updatedDiscardSave

    Senior Backend Engineer

    Senior

    GoPostgresSystem DesignObservability

    System design judgment

    WRITTEN · 30 min · 3q

    A read replica is lagging 40s behind under write load. What do you check first?

    • Replication slot / WAL sender backlog
    • Add a second read replica
    • Increase the connection pool size

    1 / 3 · 10 pts

  6. then · candidates submit
    Priya N. · 8/10 · high confidence

    Scoring each answer with self-consistency: rationale, strengths, and gaps, plus a confidence level.

    PNPriya N.high8/10

    Concrete trade-off reasoning, idiomatic Go. Caching strategy left implicit.

    Trade-off reasoning Idiomatic Go Caching depth
    Override anytime. Your call is final.
  7. 3 candidates ranked

    Ranked and ready for review. You make the final call.

    • PNPriya N.92
    • SKSam K.88
    • MLMei L.81
  8. 5 competencies · 3 assessments · 15 questions

Refine it — “make Q2 harder”, “add a coding round”…

⌘↵to send

Built to make a hiring decision you can defend.

01

Technical screening

Multiple-choice built from the job's real trade-offs — every wrong answer is a weak approach someone actually takes.

technical screening · fundamentals

Question 3 of 8 · Senior

A relational service is write-bound at peak. What is the most durable first move?

  • Add a read replica and route all reads to it
  • Shard by tenant and buffer writes ahead of Postgres
  • Increase the connection pool size
  • Vertically scale the primary indefinitely
02

Live interview

A live interview that follows up on what they say, with criteria drawn from the calls the job leaves open.

interview · live rec
AI Interviewer
Amara O.

AIHow would you scale a write-heavy service to 10× traffic?

End
03

Judgment under ambiguity

Written prompts that put candidates inside a real open decision, not a puzzle with one answer.

problem solving · response14:22 left

Question

Scale a write-heavy service to 10× traffic. Walk through your data model, indexing, and the failure modes you’d watch.

I’d shard by tenant and put a write-ahead buffer in front of Postgres, then batch-flush on a fixed interval

312 words Scored on submit
04

Evidence you can read

The verbatim quote, the finding, and where nothing was established.

scoreboard · ranked 3 scored
PNPriya N.high8/10

Concrete trade-off reasoning; idiomatic Go under time pressure. Caching strategy left implicit.

Trade-off reasoning Caching depth

Evidence by requirement

  • PNPriya N.priya.n@example.comHired92
  • SKSam K.sam.k@example.comReviewed88
  • MLMei L.mei.l@example.comAwaiting81

A real video interview that asks the second question.

interview · live
Sr. Backend Engineer rec12:48
AI Interviewer listening
Amara O. listening
End

Catch the red flags. Keep the call.

Monitoring and integrity signals run alongside every assessment — always advisory, never automated.

  • Monitored assessments: periodic screen and camera stills, on any invite.
  • Integrity signals: tab-away, paste, screenshot, and rapid answers.
  • Always advisory, never automated. The decision stays yours.
session · integrity
SKSam K.
Link · flagged
  • Tab away×2
  • Paste detected×1
  • Screenshot attempt
  • Answered in 2s

Advisory only. Flagged for review, you make the call.

Send an assessment by email, or open it to a link.

Email someone directly and they go straight in. Share an open link and you approve each person who asks.

  • Invite by email, or share one open link for a job post.
  • Approve link arrivals yourself — nobody starts unannounced.
  • Monitoring is set per role — it applies to every invite type.
share · access

Add people

amara@studio.dev devon@kit.io priya@…

People with access

PNSKML
3 invited · 2 via link

Anyone with the link

You approve each person before they start.

northassay.com/assessment/9f2c-7b1e
1 waiting for approvalApprove

From the job post to the evidence, in three steps.

  1. Paste the job

    1
    Senior Backend Engineer

    Go + Postgres, leads a small team.

    GoPostgresSystem design
    Resume a draftStart from a suggestionWritten in your own words

    — 4 open decisions found —

  2. Confirm the calls

    3
    4 open decisions
    • Ship fast or ship safe?Speculative
    • Own the on-call rota?Inferred
    • Rewrite or extend payments?In the brief
    • Mentor two juniors?In the brief
    Edit a decisionDiscardConfirm 4 decisions

    — 4 decisions confirmed —

  3. Read the evidence

    3
    Evidence by requirement
    • PNPriya N.92
    • SKSam K.88
    • MLMei L.81
    Invite by linkSend an emailRead what was established

    — end of list —

The questions you’re already asking.

Fair. It finds the calls your brief leaves open and shows you the evidence. You decide, and you can override any score.
Neither do we. Assessments are clearly authored. The video interview is opt-in and upfront.
Yours. We don’t train shared models on it. Ask us and we’ll export or delete it.
Where you hire matters. You make the hiring decision, not the model. Candidates are told: “An AI scores your answers. A person on the hiring team reviews them and makes every decision. Nothing here rejects you automatically.” NYC’s Local Law 144 also wants a bias audit you commission.
We’ve limited it, not measured it. The scorer never receives a name, photo or country, and interviews are scored from the transcript as text.
As long as the role does. They are kept for as long as the hiring team keeps your application on file, and deleted when they delete your record — the same statement the consent screen shows the candidate before anything is captured.
It tests the opposite thing. A skills test asks whether someone can do the task. This asks which call they’d make when the brief doesn’t say.
Then you’ll see exactly where. Every decision is graded by where it came from — in the brief, inferred, or speculative — and you confirm or edit each one before anything is built.
There’s a score, and it isn’t the answer. Assessments are scored, but the judgment evidence never rolls into that number — you read the quote and the finding per requirement.

See what your job description leaves open.

Paste a job post. NorthAssay shows you the calls it leaves open before it builds anything. No credit card, no sales call.

Paste a job description

NorthAssay

4 open decisions · confirmed

Not ready to sign up? Get the guide.

The 5-minute version: five checks that separate real skill from a good profile.

NorthAssay

AI hiring assessments, scored with rationale you can defend.

Guides

All screening guides

Assessment methods

Candidate assessmentSkills assessmentsTechnical skills assessmentCommunication assessmentPre-employment assessment

Tools & software

Assessment tools comparedAI candidate screeningHiring software for small teams

By role

Technical interview questionsAssess a DevOps engineerAssess a data engineerAssess an engineering manager

Screening practice

Five checksWhen candidates use AI

Product

How the product worksVerified answersHow it worksJudgment evaluationAI video interviews

Get started

Start freePricingContactSign in

Trust

Responsible AISecurityAbout
© 2026 NorthAssay. All rights reserved.
PrivacyTerms