Know who's exceptional, and who's a risk, before you meet them.
Kenning interviews every candidate in depth and gives you a one-page report, with every judgment backed by their own words.
Candidate 0418
5 of 10 core abilities are clear strengths. No concerns, and the answers are consistent.
You said static analysis gives up precision to stay computable. In your null-pointer checker, where did you give it up, and why there?
Path sensitivity. On 4 of the 15 projects the solver blew our 30-minute budget…
Sample data from a PhD admissions interview. It ran in Mandarin and is shown here in English.
For any role where thinking matters
- PhD admissions · Computer science
"Your resume says the checker scales to a million lines. What's the largest codebase you actually ran it on?"
Depth of expertiseSelf-knowledge - Software engineering
"Tell me about an outage you caused. What did you change afterwards?"
OwnershipCoping with setbacks - Product management
"Two teams need the same engineer next sprint. How do you decide, and who do you tell first?"
Decision-makingReading people - Data science
"Your model's accuracy jumped 12 points overnight. What do you check before you believe it?"
Analysis and reasoningSelf-knowledge - Consulting
"Roughly how many coffee shops are there in Shanghai? Think out loud; it doesn't need to be exact."
Analysis and reasoningCommunication
- Investment analysis
"Which number in that pitch deck would you check first, and why that one?"
Seeing beneath the surfaceDecision-making - Research labs
"What did you stop doing so you could finish that paper on time?"
PlanningFollow-through - Fellowships and scholarships
"Tell me about something you learned that nobody asked you to learn."
LearningValues - Startup programs
"Your co-founder disagrees with you in front of the team. What happens next?"
CollaborationLeadership - Teaching
"A student keeps missing deadlines and won't say why. What do you do this week?"
Reading peopleView of people
Sample questions. Field knowledge comes from a skill pack; program analysis is in use today, and packs for other fields are drafted for your review.
How it works
- 01
Send a link
Add the candidate's resume and send them a private link.
- 02
They chat for 20 to 30 minutes
Questions adapt to every answer. No scores, no timers.
- 03
You read one page
Every judgment links to the quote behind it. You make the call.
The whole person, not just their expertise
27 abilities in five groups. Most of them are things a resume can't show.
Knowledge
- Depth of expertise
- Breadth and transfer
- Learning
Information processing
- Breaking things down
- Summarizing
- Matching problem to method
- Finding answers
- Generating ideas
Thinking
- Seeing beneath the surface
- Analysis and reasoning
- Decision-making
- Planning
- Communication
- Self-knowledge
Worldview and values
- Worldview
- View of people
- Values
Character
- Interests outside work
- Coping with setbacks
- Managing energy
- Reading people
- Composure
- Leadership
- Follow-through
- Presentation
- Ownership
- Collaboration
Bold: the 10 core abilities that decide the result.
Questions they can't rehearse
Every question is written for this candidate from their resume and their last answer. There's no question bank to leak.
Question
You said static analysis gives up precision to stay computable. In your null-pointer checker, where did you give it up, and why there?
- Builds on
- "you give up precision to get something you can actually compute" turn 2
- Looking for
- Depth of expertiseDecision-making
- Strong answer
- Names the trade-off they made and the measurement behind it.
- Weak answer
- Repeats the textbook definition of path sensitivity.
Resume claims get checked
Big numbers get followed up in the conversation. If an answer doesn't match, the report says so.
| On the resume | In the interview | Status |
|---|---|---|
| Found 42 bugs, 23 confirmed by developers | Named the projects and how reports were triaged. turn 11 | Verified |
| Scales to million-line codebases | Largest run was about 200k lines; the 1M was an estimate. turn 7 | Doesn't match |
| Evaluated on 15 open-source C projects | Gave build details; number not yet checked. | Has details |
Find the outliers on both ends
Kenning flags the clearly exceptional and the seriously concerning, and marks close calls for a second look. Nothing is final until you confirm it.
- Turned hostile when asked about an inconsistency in their own account. turn 16
- Backed by a contradiction about their role turns 4–5 and no detail after a direct follow-up.
Close to the line: one more clear strength would make this "Likely exceptional".
Built on research, not gut feel
Structured interviews predict job performance about twice as well as unstructured ones.
Sackett et al., 2022 [1]Interviewers feel just as confident when the answers are random. So every judgment shows its quote.
Dana, Dawes & Peterson, 2013 [2]People spot lies barely better than chance. So Kenning checks facts, not faces.
Bond & DePaulo, 2006 [3]
Questions
How long does an interview take?
Usually 20 to 30 minutes. It stops once the core abilities are clear, and you can change the length.
Can candidates prepare, or paste answers from an AI?
There's no fixed question list to prepare for, and follow-ups ask for specifics only someone who did the work would know. Answers that look pasted are flagged.
Does Kenning decide who gets in?
No. Every result starts as unconfirmed, with the quotes behind it. You confirm or overturn it.
Can we change what it looks for?
Yes. Describe what matters to your group in plain words; Kenning drafts the change and you approve it. Every version is saved, so past reports can be rebuilt exactly.
Which fields and languages does it support?
Any role where you care how people think. Field knowledge comes from a skill pack; program analysis is reviewed and in use today, and Kenning drafts packs for other fields for your review. Interviews currently run in Mandarin.