AI video interviews. Human hiring decisions.
Interview every candidate. Decide on the evidence.
Bibha Meet runs your first-round interviews over video. Its AI interviewer asks your questions, runs coding tasks, checks that the same person answers throughout and writes a scored report that quotes the candidate's own words. Your team reviews it and makes the call.
What does Bibha Meet do?
Bibha Meet is an AI interviewer for recruitment agencies. It interviews candidates over video, watches for stand-ins and outside help, can check a photo ID against a selfie, grades coding work and writes a scored report. Recruiters and hiring managers read the report and replay the moments that matter, and a hiring manager or agency admin records the decision. Candidates join from a private link, with no account to create.
Before the first round
Can you trust a first round you did not sit in on?
Each recruiter now handles far more applications than in 2021, with fewer colleagues to share them. Video calls have also made it easier to send a stand-in or read out an answer someone else wrote.
A shortlist built that way is hard to defend when a client asks why a candidate was put forward. Bibha Meet gives every candidate the same structured interview and keeps the evidence behind each score.
93%
more applications per recruiter than in 2021, while recruiting teams shrank by 14%.
Gem, 2026 Recruiting Benchmarks (opens in a new tab)35%
of hiring managers say someone other than the listed applicant has taken part in a virtual interview.
Checkr survey of 3,000 managers, 2025 (opens in a new tab)46%
of US hiring managers have seen interviewees use AI to answer interview questions.
Resume Genius, 2026 (opens in a new tab)
From job description to decision.
Your team sets up a role once and comes back to finished reports. Candidates take the interview when it suits them, evenings and weekends included.
Build a template
Paste the job description. The AI suggests the skills to test and drafts questions and coding challenges. You keep or drop each draft and can edit it afterwards.
Invite candidates
Send a private link to one candidate or a whole batch. They start any time within seven days, or at a slot you book.
The AI interviews
Candidates run a short system check, give consent and talk to the AI interviewer. Coding tasks open in the same session.
Review and decide
Scoring starts when the interview ends. Your team reads the report, checks any flags against the recording and records the decision.
An interviewer that keeps to your brief.
The AI interviewer opens by telling candidates it is an AI and that a person makes the hiring decision.
Then it works through your questions in order and probes wherever an answer stays thin.
Follow-ups that ask for specifics
When an answer stays general, it asks for one concrete detail: a number, a decision or the part the candidate did personally. It avoids generic prompts like 'tell me more'.
Neutral from start to finish
It does not praise answers or hint whether they were right. The last candidate of the day gets the same tone as the first.
Time to think
Candidates can ask to hear a question again, say they are still thinking or ask for quiet while they code. If the connection drops, it waits up to 5 minutes and repeats the question that was cut off.
Ten languages
It interviews in English, Hindi, Spanish, French, German, Brazilian Portuguese, Italian, Dutch, Russian and Japanese, with Indian and UK English as options. Hindi interviews follow answers that mix in English. Candidate screens and emails are in English.
Let the AI lead, or lead it yourself.
Some clients want a person in the room. Both ways produce the same scored report and cost the same per minute.
AI-led
The AI runs the whole interview, coding included. Nobody from your team has to be online while it happens.
Recruiter-led, with a copilot
Your recruiter asks the questions while a silent copilot suggests up to two follow-ups after each answer, based on what the candidate said or left out. Only the interviewer sees them.
Watch and steer
Open any AI-led interview while it runs, follow the transcript and integrity alerts as they happen, and type a question for the AI to ask at the next pause.
Know who answered, and how.
In AI-led interviews, proctoring watches the camera, the microphone, the shared screen and the browser. Recruiter-led interviews keep the browser checks.
Each finding is saved against the moment it happened, and camera findings keep a snapshot.
Identity at the door
Switch on a photo ID and selfie check with random movement prompts. The face is matched to the ID and the name is read from it. Borderline matches go to a recruiter instead of failing the candidate.
The same person throughout
The face in the seat is compared with the face that started the interview, or with the ID selfie. A change of person raises a serious flag, and once a reviewer confirms it, the stand-in's answers stop counting.
A phone or a second person
The camera looks for a phone, a book, a second laptop, a TV screen or another person in view. An optional phone side camera shows reviewers the desk.
Voice checks on every answer
Each answer is checked against the candidate's own voice, and the system notes when a different voice answers right after a question.
A follow-up when an answer sounds borrowed
If an answer shows signs of outside help, the AI asks one ordinary question about a detail of the candidate's own work. It never mentions monitoring.
Screen and browser
Candidates share their whole screen. Tab switches, pastes from outside the code editor and developer tools are recorded. A short slip pauses the interview with a countdown, and only a long absence ends it.
Flags pause a decision. People make it.
No flag rejects a candidate. A serious one holds the AI recommendation until someone on your team has looked at the evidence.
A score that shows its working
Integrity starts at 100 and loses points by severity, with caps on repeated events. Below 85 the recommendation is held. The report lists every event, says whether it counted and states which checks ran.
A hold, then a recorded decision
A recruiter or agency admin clears or confirms the concern. Hiring over a confirmed concern needs a written reason, and every review and override goes into the audit log.
Fair to honest candidates
Looking away to think costs nothing on its own, and dropped connections are never charged. Approved accommodations switch off the checks that would misread a candidate. There is no emotion analysis.
Coding inside the conversation.
Technical candidates write and run code during the interview instead of on a separate test link. The AI interviewer follows the code as it is written, states the result and asks about the approach.
Optional settings give each candidate a different version of a challenge, or allow an AI assistant in the editor and score how the candidate uses it.
Programs in 7 languages
JavaScript, TypeScript, Python, Java, C++, Go and Rust. Candidates run the visible examples, then submit against hidden tests that report only pass or fail.
Projects for front-end and full-stack roles
Candidates build or fix a multi-file React, Vue, HTML and Sass, or Node.js project with a live preview and automated checks.
An isolated sandbox
Every run happens in an isolated sandbox with no internet access and no secrets, reset before each run.
Questions about the code
After each submission the AI asks a review question. Once every test passes, it asks how the solution would cope with a changed constraint.
Challenges drafted for you
Describe the role in a sentence or paste the job description. The AI drafts a challenge with starter code and hidden tests, and keeps only the tests its reference solution passes.
Evidence for reviewers
Reviewers see every run, the share of pasted code and a replay of the typing on program challenges. Code that closely matches earlier candidates in your workspace is shown as context and never changes a score.
A report you can check line by line.
Scoring starts when the interview ends. The report uses only what the candidate said and submitted, and each question's analysis quotes the transcript.
Five scores and a capped verdict
Technical skills, communication, problem solving, cultural fit and experience, each from 0 to 100. The overall score caps the recommendation, so a weak score cannot carry a strong verdict.
Blind to names and schools
Before scoring, the candidate's name and, where found, employers, schools and places are swapped for placeholders. The scorer is told that accent, fluency and pace never count. When a role asks for an English level, it is estimated separately and never changes the scores.
Evidence for each question
Each question gets its own analysis, with quotes checked against the transcript and a list of claims worth confirming with references.
Too little to judge is not zero
When a candidate says too little to assess a skill, the report marks it not assessed instead of scoring it zero.
People decide
Reviewers score without seeing each other's marks until they submit their own. A hiring manager or agency admin records hire, hold or reject.
Ready to share with a client
Integrity events open the recording at the right moment, and the report downloads as a PDF with your agency's logo.
Built around how agencies work.
Each agency works in its own workspace, and one sign-in opens every workspace a person belongs to.
Roles for each side
Agency admins run the workspace, recruiters run interviews, and hiring managers see only the candidates assigned to them.
Your brand on the report
Add your logo, and an accent colour for the console. The report PDF carries your logo, and candidate emails carry your company name.
Candidates from your ATS
Import candidates from Greenhouse or Lever, upload a CSV, or add them one by one. In TalentFlow, candidates who clear the screening call move straight on to a Bibha Meet interview.
Analytics that grow with you
After five completed interviews, see completion, pacing and interviewer quality. Record outcomes to compare recommendations with who was hired.
Pay for minutes, not checks
One credit is one interview minute, charged for the minutes used. Identity checks, proctoring, coding and the report add nothing.
Nothing wasted
Cancellations, no-shows and expired links cost nothing, approved extra time is not billed, and credits never expire.
Privacy and security
What we keep, and what we delete.
Candidates are told what is recorded before they start. The app, its files and logs run on AWS in Mumbai, and live video, speech and AI processing use outside processors under contract. Ask us for the processor list and a walkthrough of the controls.
Consent on record
Saved with the time, the IP address and the wording the candidate agreed to.
Recordings expire
Deleted after 90 days by default, or on the schedule your agency sets.
Face and voice references removed
The references built during an interview are deleted about 5 to 10 minutes after it ends. ID check photos stay until the candidate is deleted.
No emotion analysis
Nothing reads emotions, stress or personality. Scores come from what the candidate said and submitted.
Access is logged
Recording links, report downloads, ID images and live observers are written to the audit log.
Protected sign-in
Sign-in is rate limited, sessions end after 8 hours without use, and each user can turn on two-factor sign-in.
Start with one open role.
Tell us about the role and how many candidates you expect. The Bibha team sets up your workspace, and you can send the first invitation as soon as your template is ready.
Questions and answers
What agencies and candidates ask us most. For anything else, write to support@bibha.ai.
Does the AI decide who gets hired?
No. The AI scores the interview and makes a recommendation. A hiring manager or agency admin records every decision. In AI-led interviews, the AI tells candidates this at the start.
What do candidates need?
A laptop or desktop with a camera and microphone, a current browser and a steady connection. They open the link from their email, with no account and nothing to install. Phones and tablets cannot share a whole screen, so proctored interviews need a computer.
Can it catch every kind of cheating?
No tool can. Bibha Meet checks the camera, the voice, the shared screen and the browser, and asks follow-up questions when an answer looks borrowed. It cannot see a phone held out of camera view, and it does not detect deepfakes. Whatever it finds goes to a person to review.
What happens if the connection drops?
The interviewer waits up to 5 minutes. When the candidate is back, it welcomes them and repeats the question that was cut off. If they do not return in time, the interview ends and the score reflects the questions it reached. Disconnects never count against the integrity score.
Can candidates ask for adjustments?
Yes. Before the interview, a candidate can ask for extra time, more time to speak, approval to bring their own interpreter, permission to use a hearing aid, or eye and head direction checks switched off. A recruiter approves each request, and the matching checks stay off for that candidate.
How is it priced?
One credit buys one interview minute. Scheduling holds the template minutes, the charge is the minutes actually used, and the rest goes back. Identity checks, proctoring, coding and the scored report use no extra credits, and credits never expire.
Does it work with our ATS?
Bibha Meet is integrated with TalentFlow, Bibha's recruitment platform. TalentFlow screens candidates on a call, and those who move on take their video interview in Bibha Meet. Bibha Meet also imports candidates from Greenhouse and Lever, but results are not written back to them. Other ATS vendors can connect through our partner API.
Where is candidate data kept?
The app, its files and logs run on AWS in Mumbai. Live video, speech and AI processing run at outside processors under contract, and we share the full list in a security review.
I am a candidate. Where do I sign in?
You do not need to. Open the interview link from your email. It works without an account, and the page walks you through a short system check first.
