AI Interview· Screening interviews, run by AI
From a job description to a decision-ready report.
Vriti's AI interviewer runs a structured, job-specific voice interview — with live coding when the role needs it — records and proctors it, and hands your team a scored, evidence-backed report.
- You set the bar. AI drafts the criteria and questions from your JD; you edit the weights, the questions and the cut-off.
- Candidates just join. No account needed — a secure link, a device check, and a real-time conversation with the AI.
- Every answer is evidence. Scores by criterion, AI reasoning, transcript, recording and an integrity verdict in one report.
1Create the role· Step 1 of 6
Start with the job you're hiring for.
Upload a JD, write one, or have AI draft it from a one-line prompt. When you continue, Vriti reads it and pre-fills the interview: title, experience, level and the criteria worth scoring.
- Job Description
- Preferences
- Evaluation Criteria
- Objectives
Drag and drop your JD here, or browse
PDF, DOCX or TXT · up to 100 MBYour generated JD appears here. You can edit it, or prompt again — each result replaces the last.
- Three ways inUpload a PDF, DOCX or TXT, write in the editor, or generate and refine with AI as many times as you like.
- Structured on arrivalThe job title, experience range, suggested criteria and interview defaults are extracted from the JD for you.
- Jobs you can manageEvery job lands on your Job Postings board — search, clone, disable, assign to teammates or export the list.
Next step getsSenior React Developer · 5–8 years · 5 suggested criteria
2Prepare the interview· Step 2 of 6
Decide how the AI interviews — and what it scores.
Pick the interview type, level, length and cut-off. The AI drafts weighted qualifying criteria and a full interview plan; you tune both. Try changing the type — the criteria are regenerated, and the rest of this page follows your settings.
15 of 45 minutes go to coding.
Qualifying criteria
- Component and state design≥ 75%Splits state by who reads it; knows when context stops scaling.
- Rendering performance≥ 70%Finds wasted renders with a profiler before changing code.
Qualifying criteria generated.
Then the AI builds the interview itself.
Objectives, sub-objectives and the questions the AI will ask — including coding problems — each with what it's evaluated on. Every edit saves as you make it.
1.1React rendering and state
1.2TypeScript at scale
2.1Live coding
3.1Project deep-dive
- Conversation, code, or bothOne configurable AI interview per job — set the level, duration from 30 to 60 minutes or custom, and how many coding problems.
- Criteria you can defendWeighted criteria with indicator thresholds, drafted by AI from the JD and fully editable, or regenerated in one click.
- Your voice, your languageChoose the AI engine, the interview language and the voice, and hear a sample before candidates do.
Next step getsBoth · Senior · 45 min · Cut-off 70% · Cost-optimized
3Invite the candidate· Step 3 of 6
Send a link. Or let them pick the time.
Add candidates by name and email — a resume is optional and parsed into a profile. Send a link that stays open for a window you choose, or turn on slot selection and let each candidate book their own start time.
- Many candidates, one stepInvite several people at once, with a custom subject and CC recipients. Vriti sends the emails.
- Self-schedulingCandidates pick a start time in their own time zone, with 1 to 5 attempts to change it. A missed slot uses an attempt and triggers a follow-up.
- Change of plansReschedule or cancel in bulk from the job, or give selected candidates a common new time slot.
Next step getsAarav Mehta · join link open until Wed 30 Sep, 10:00
4The candidate interviews· Step 4 of 6
A real conversation, with an interviewer that never tires.
Aarav checks his camera and mic, reads the guidelines and shares his screen. Then the AI interviews him by voice — following your plan, asking follow-ups, and opening a coding problem when the job calls for one. Try the buttons under the room.
1Device check
- CameraOK
- MicrophoneOK
- SpeakerOK
2Guidelines
Six do's and six don'ts, and a notice: video, audio and screen signals are used, and sessions are recorded and may be reviewed.
3Share entire screen
Sharing the whole screen and going fullscreen are required to join.
The room opens once the device check, guidelines and screen share are done.
- Resilient by designA heartbeat every two seconds, a 240-second reconnect window with the timer paused, and a new link by email if the interview can't recover.
- Live coding, in the roomThe AI opens an editor with the problem, hints and a timer. 11 languages, skip and return, auto-submit when time runs out.
- Recorded and proctoredScreen plus both voices are recorded. Webcam frames and browser events feed the integrity analysis. Candidates join from a desktop browser.
Next step getsRecording, transcript and proctoring signals · plus the submitted code
5The AI evaluates· Step 5 of 6
Every score traces back to something the candidate said.
The AI scores each answer and each line of code against your criteria, then weighs them into one score out of 100. Pass or Fail follows your cut-off — move it and watch the decision change.
EvidenceFrom the interview
I recorded it in the React profiler. Every quantity change re-rendered all the product cards…
I'd move the cart into a store with selectors, so each card subscribes only to its own line item.
I'd pass the row type as a generic and type the columns from its keys.
Sorts by start time, then merges in one pass. O(n log n).
Yes, it mutates their data. I'd copy it first, which keeps the same complexity.
CriteriaScore / weight
Overall
88/100Strong29 + 14 + 22 + 14 + 9 — weighted by your criteria.
DecisionAutomatic
- Shortlisted — strongPass, 85+
- ShortlistedPass
- Rejected — borderlineFail, 45+
- RejectedFail
Integrity is analysed separately from the score — it tells your team what to review, not whether the candidate is capable.
- 0Multiple faces
- 1Left frame
- 1Tab switches
- 0Camera stops
- 3Eye movements
- 2Head movements
- Weighted, not averagedCriterion and objective weights decide how much each part counts, so the score reflects what the role needs.
- Explained, criterion by criterionEach score carries the AI's reasoning, plus the strengths and areas for improvement it found.
- A screening recommendationShortlisted or Rejected tells you who to move forward. The hiring decision stays with your team.
Next step gets88/100 · Pass · Shortlisted — strong · integrity verdict
6The report· Step 6 of 6
Hand your team a report they can act on.
Everything from the interview in one place: the score and why, every criterion, every answer, the code, the recording and the integrity analysis. Jump from any line of the transcript to that moment in the recording.
Aarav MehtaPassed
Strong frontend depth with measured, evidence-led answers. Solved the coding problem with a correct O(n log n) approach and handled the follow-up well. TypeScript is the weakest area for a senior hire.
- 3Objectives
- 5Criteria
- 7Evidence
- 6Questions
- 1Coding
Key strengths
- Diagnosed a real performance problem with a profiler before changing code
- Clear trade-off reasoning on where state should live
- Asked a clarifying question before writing any code
Areas for improvement
- Missed the empty-input case until prompted
- Answers on TypeScript generics stayed at surface level
Score by dimension
Objectives
- Built for reviewAn evidence explorer filtered by objective, coding or verbal, and score band; a searchable transcript with per-answer signals.
- Recording with contextPlay at 0.5–2×, skip ±15 seconds, and see proctoring moments marked on the timeline.
- Ready to shareSummary, Detailed and Transcript PDFs, plus sharing by email or WhatsApp. What each person sees follows their plan's permissions.
Next step getsA decision-ready report for your team — exported as PDF, or shared in two clicks
Across every role
Aarav's interview is one of many.
The dashboard adds up every interview across your jobs, your team's and your organisation's. Access follows roles and plan permissions, and the lifecycle runs without anyone chasing it.
Overview
- Total interviews31
- Completion rate71%
- Avg AI score74 /100
- Time saved16.5 hEstimate · 45 min per completed interview
Interview activity
Hiring funnel
- Scheduled31
- Completed22
- Shortlisted15
Team & access
Invite recruiters by form, Excel upload or HR-system import, set reporting lines, and assign jobs to teammates.
- RecruiterMy JobsCreates jobs, schedules interviews and reads reports.
- Manager+ My Team JobsSees the jobs of everyone who reports to them.
- Super admin+ Org JobsSees every job, and adds or deactivates recruiters.
Turn it off and the average score on the dashboard is masked — features follow your plan.
The lifecycle runs itself
- ScheduledThe invitation goes out when you schedule.
- In progressStarts when the candidate joins inside the window.
- CompletedThe recording is saved and the evaluation begins.
- Pass or FailSet by the AI evaluation against your cut-off.
Works with the systems you already use
- Inside your ATSEmbed the recruiter experience in your ATS with your own sign-in. Connector setup is done with Vriti.
- Results by signed webhookInterview results are pushed to your system with an HMAC signature, configured with the Vriti team.
- Embeddable widgetsDashboard, scheduler, interview and report widgets for your own pages.
- Numbers you can filterDate presets or a custom range, a per-job filter, and export to PDF, Excel or CSV — or email the report.
- The whole pipeline in one viewEvery interview with status tabs, search and date filters, and a calendar of scheduled interviews and candidate-picked slots.
- Honest measuresTime saved is an estimate — 45 minutes for each completed interview — and it's labelled that way.
Let the AI run your first-round interviews.
Set up a job in minutes, invite your candidates, and review scored, evidence-backed reports — while your team spends its time on the people worth meeting.
- Create the role
- Prepare the interview
- Invite the candidate
- The candidate interviews
- The AI evaluates
- The report