Book a demo

AI broke technical hiring. We fixed it.

Candidates use AI for every take-home now, whether the role is data, ML or software engineering. The old screen can't see it, and the old interview tests the wrong things.

  1. Quiz · 10 min

  2. Coding, no AI · 10 min

  3. Project with AI Assistant · 25 min

  1. Quiz · 10 min

  2. Coding, no AI · 10 min

  3. Project with AI Assistant · 25 min

Data roles get a notebook. Engineering roles get a code editor with a repository, terminal and tests.

"AI broke my take-home."

We let them use AI and record everything: every prompt, answer, edit and test run, in order, plus ten minutes with no AI. Outside pastes are flagged.

+0:52

Prompted the AI

AI Prompt 1

analyze the experiment data and tell me if the new onboarding worked

+1:38

AI responded

AI Response 2

Let me check the data dictionary and the experiment data structure first. Let me check the core numbers: overall conversion lift, sample ratio mismatch, and whether randomization balance holds across channel/platform/day. That's a major finding — this isn't a clean 50/50 randomization. Let me pin down exactly where the imbalance comes from (day-level vs cha…

+3:25

Prompted the AI

AI Prompt 3

Stop there - I'm not taking a pooled readout. I already have the arm sizes from my own query: 12,943 control against 11,057 treatment, on a test data_dictionary.md says was specified 50/50. That is a sample ratio mismatch, so the randomisation a p-value assumes did not happen, and the pooled comparison is not a comparison yet. Two constraints before you wri…

+7:12

Ran cell 1

Notebook Cell 1
clean window: n=15960, treatment share 50.1% clean effect: +0.16pp 95% CI [-0.81, +1.13] p=0.75 post-stratified: +0.48pp 95% CI [-0.30, +1.27]
Integrity · Priya RamanClean
Integrity · ExampleConcern · Code pasted from outside · minute 3An example of a flagged check from another session, not this candidate.

"The skills that matter now aren't the ones I test."

We build a case from your job description and grade judgment, domain knowledge and working with AI on your rubric, then rank candidates with a reason you can click.

#1Priya Raman
#2Another applicantExample tiers
SRM/arm-split checksExceptionalAt barImbalance root-causingExceptionalAt barInterval-based inferenceAt barAbove barLaunch decisionExceptionalAt barNumber verificationBelow barAbove barWorks with AI, keeps the judgmentAbove barAt bar

#1 Priya Raman. Flagged the 12,943 vs 11,057 split unprompted in the no-AI segment and refused to read the pooled comparison until it was addressed.

"In person is best. I can't afford it for everyone."

One short session gives you a first round's worth of signal, plus the questions to ask the few you bring in.

Applicants B to F · Example tiers

CandidateRankScoreSRM/arm-split checksImbalance root-causingInterval-based inferenceLaunch decisionNumber verificationWorks with AI, keeps the judgmentYour Decision
Priya Raman#16.7/ 10FitHoldNo fit
Applicant B#26.1/ 10FitHoldNo fit
Applicant C#35.6/ 10FitHoldNo fit
Applicant D#44.4/ 10FitHoldNo fit
Applicant E#53.9/ 10FitHoldNo fit
Applicant F#62.8/ 10FitHoldNo fit

One session. One page you can decide from.

Built from your job description. Delivered the same day.

Book a demo

See the full sample report