Catch proxy interviews and deepfake candidates before the offer.
Your background check runs after the offer, by which time a proxy has already sat the interview. Berri360 runs your first-round technical interviews itself, so it can check for the whole session that the person answering is the person you invited.
Running since 2019 and in use at four of India's largest IT services firms.
Used by teams at Fortune 500 companies across India, North America, LATAM, SEA and more
Why background checks miss interview fraud.
In most hiring processes the identity check is the last step, after the application, the résumé screen, the coding test, the interview and the offer. By the time it runs, the person already has a signed offer and a start date, and the check itself relies on documents they handed over.
of background checks in Indian IT and tech services turn up a discrepancy, after the offer.
of hiring managers say someone other than the applicant took part in a video interview.
is how often untrained people correctly spot a deepfake, which is barely better than a coin toss.
An AI interviewer that checks who it is interviewing.
Berri360 covers the first half of hiring in one system: it ranks the applications, contacts the candidates and runs the first-round technical interview with a real coding exercise, checking identity the whole time.
Where a verification tool bolted onto a video call checks a face at the door and then stops looking, Berri360 stays with the candidate from the first question to the last, because it is the one asking them. It is the interview layer of BerriBot's agentic AI recruiting platform.
- 01Screening
- 02Outreach and scheduling
- 03The interview
- 04Identity checks
- 05Handover to your ATS
Everything runs in one system, under one contract, on one candidate record.
Six checks that catch proxies, deepfakes and cheating.
These checks run for the whole interview instead of once at the door, and every flag arrives timestamped, with what triggered it and how confident the system is. For the detail behind each one, see Berri Proctor or our guide to deepfake and proxy candidate detection.
Liveness
Confirms a real person is on camera, so a photo, a looped video or a replayed recording fails at once.
Face match
Keeps checking that the face on screen belongs to the person who started the interview.
Voice match
Compares the voice with itself for the whole session, so a swap at minute 20 is flagged when it happens.
Lip sync
Catches answers fed from off camera, because the mouth and the audio drift apart when someone else is speaking.
Gaze and head
Measures where the candidate is looking, since reading from a hidden screen leaves a pattern a person would miss.
Screen and room
Watches for tab switching, screen sharing, pasted answers, second monitors and a second person in the room.
Each flag is evidence for your team to review, and the decision stays with them.
Berri MasterMind runs the first round the same way every time.
The AI interviewer asks about the projects on the CV, follows up on what the candidate actually said and then sets a real coding problem in the browser, so you get a scorecard with the transcript quoted under every score.
- 140+ languages
- Same rubric for every candidate
- Your interviewers meet a checked shortlist
Berri Connect calls from a number candidates trust.
Berri360 reaches candidates by phone, WhatsApp, SMS and email, books the interview and sends the reminders, and because calls show as Truecaller-verified, people answer them. When a role needs it, spoken English can also be graded on the CEFR scale.
- Phone, WhatsApp, SMS, email
- Truecaller-verified calls
- Books and reminds on its own
Berri Search & Match ranks shortlists on what a CV means.
Starting with the candidates already in your ATS, it reads each CV against what the job actually needs and scores every requirement, using matching work from a founder who holds five granted US patents in the field.
- Starts with your own ATS
- A score per requirement
- Five granted US patents
Why candidates walk out of AI interviews.
Although 38% of job seekers have abandoned a hiring process because of an AI interview, the reasons they give are fixable: being watched without being told, not knowing that AI was scoring them, and a bot that never listens.Greenhouse, 2026
Told before it starts
Candidates know that an AI runs the interview and that identity checks are on before they begin.
A real conversation
Follow-up questions build on what the candidate actually said.
Their language, their needs
Interviews run in 140+ languages, with adjustments on request.
A person decides
Someone on your team reviews every flag before it affects a candidate.
Berri360 never reads emotion, personality or mood from a face or voice, and since February 2025 the EU has prohibited emotion inference at work, a line that hiring sits inside.
Candidate data, security and compliance.
A candidate's face, voice and ID are among the most sensitive data you will hold about someone who doesn't work for you yet, so here is where that data goes and how decisions about it are made.
Where the data lives
- Data residency stated in writing, India included
- Encrypted in transit and at rest
- Retention schedule with logged deletion
- Zero-retention terms with our model providers
- No cross-company model training
How decisions are made
- The same written rubric for every candidate
- The transcript quoted under every score
- Human review before any decision against a candidate
- A full audit trail for every interview
Who Berri360 is for.
- You hire 500+ technical roles a year (it works best above 2,000)
- Your interviews are remote and will stay remote
- A bad hire lands on a client's project, not just an internal team
- Your security team reviews every vendor that touches candidate data
Not the right tool if…
- You want emotion, personality or mood scoring, which we don't build
- You want to replace your coding tests, since Berri360 works best alongside them
Proxy interview detection, answered.
What is a proxy interview?
A proxy interview is one in which someone other than the applicant answers the questions, either by sitting the interview in their place or by feeding them answers from off camera. Because most identity checks happen after the offer, a proxy who performs well can reach a signed offer before anyone notices.
How does Berri360 detect proxy interviews and deepfake candidates?
Berri360 runs the interview itself, so it checks identity for the whole session rather than once at the start. It confirms the candidate is live, keeps matching their face and voice, compares lip movement with the audio, measures gaze and watches the screen and room, then flags each event with a timestamp, what triggered it and how confident the system is.
Can it catch interview cheating, such as reading answers from another screen?
Yes. Gaze and head tracking picks up the pattern that reading from a hidden screen leaves, and the screen and room checks catch tab switching, screen sharing, pasted answers, second monitors and a second person in the room.
We already proctor our coding tests. Why do we need this?
Test proctoring protects the test, but the person who passes it is not necessarily the person who sits the interview, and the interview is where your team forms its judgment about a stranger on a video call.
What if it flags a legitimate candidate?
A flag is evidence rather than a rejection. Your team sees the timestamped clip, what triggered it and the confidence behind it, and then decides what it means.
Can a recruiter override a score or a flag?
Yes. Berri360 provides evidence and decision support, while the hiring team makes the final decision and can overrule any assessment.
How do candidates know AI is being used?
Before the interview starts, candidates are told that an AI is running it and that identity checks are on, and they are given a route to ask for adjustments or a human interviewer.
Does it write into our ATS, or send a PDF?
It writes structured data into the right fields in Workday, SAP SuccessFactors and Oracle, and we show this live against a test environment on the first call.
Can we prove it on our own data first?
Yes. Most teams start with a paid pilot on one job family of 200 to 500 interviews, with the success measures agreed before it begins: interviews flagged, candidates who finished, recruiter hours returned and what candidates said.
“Confirming that a face is live, and belongs to the same person for a whole conversation, works reliably. Reading emotion from a face does not. We built Berri360 to run the interview, because checking someone once, at the start, is not really checking.”
Raja spent three years at Meta and two at SAP building matching and computer vision systems, and holds five granted US patents on matching.
Watch Berri360 catch a proxy candidate.
In 30 minutes, with one live job description, we run a real first-round interview and show you what the checks produce, including what a coached proxy looks like when it's caught.
What usually follows is a paid pilot on one job family of 200 to 500 interviews, with the success measures agreed up front.