NewAarohi-1 full-body preview: gesture, posture and presence.Learn more →

Every candidategets a realinterview.

Aarohi is a lifelike AI interviewer. She runs structured, face-to-face first rounds around the clock, scores every answer against your rubric, and sends the evidence to your ATS.

Try the interview

Designed to sync with

GreenhouseLeverAshbyWorkdaySAP SuccessFactorsWebhooks
Live · First round06:41
You

Aarohi

Real-time render of Aarohi. The conversation is a scripted preview; try the real flow in the live demo.

The first round is broken

Résumés filter.Phone screens drift.Panels disagree.

Most candidates are judged on keywords before anyone speaks to them. The ones who get a call meet a tired recruiter on their eleventh screen of the day. Aarohi gives every applicant the same thoughtful conversation, at 3pm or 3am.

Iris reads engagement and composure, Cadence keeps turn-taking natural and asks real follow-ups, and every answer is scored against your rubric, then delivered as structured data through one webhook.

Question 3 of 6

Follow-up generated

“You mentioned the launch slipped. What would you change if you ran it again?”

24/7

Interviews in every time zone, with no scheduling back-and-forth.

1rubric

Applied identically to every candidate, every time.

<600ms

Target response time per turn, so it feels like a conversation.

100%

Of interviews transcribed, scored and searchable.

Platform design targets for the ManavOS beta.

The ManavOS models

Four models.One presence.

An AI human is more than a talking head. Each layer is built in-house and tuned to work together, with Claude as the reasoning layer by default, or your own LLM.

Iris

Perception

Reads expression, gaze and on-screen context in real time, so the interviewer notices hesitation and adapts.

Lumen

Rendering

Photoreal faces and, in preview, full bodies: micro-expressions, posture and gesture rendered at video frame rates.

Cadence

Conversation

Voice, tone and turn-taking. Knows when you've finished a thought, and when you've only paused to think.

Recall

Continuity

Remembers people across sessions, so a coach or tutor picks up exactly where you left off.

Explore the models →

Live demo

Sit the interviewyourself.

Pick a role and answer four questions. Watch Aarohi listen, follow up, and score you in real time against the same rubric every candidate gets.

Ready
Aarohi · AI interviewer

Rubric

Demo scoring

Structure–
Evidence & metrics–
Ownership–
Clarity–

The transcript appears here. Each answer is timestamped and linked to the rubric line it supports.

Preview build: questions and scoring run locally on mock data. Production connects real-time video and voice, with Claude reasoning over your rubric.

For teams that build with AI

Your hiring datais only as good asthe interview behind it.

Inconsistent interviews produce noisy data. Aarohi asks every candidate the same core questions, follows up where it matters, and scores against one rubric, so the signal you train and decide on is clean.

Each session returns a full transcript, perception notes and a structured scorecard, delivered to your ATS, an S3 bucket or any webhook within seconds of the call ending.

Candidate
0412 · Data analyst
Overall
3.4 / 4 · Advance
Structure
3.6 · Clear situation → action → result
Evidence
3.1 · Quantified impact in 3 of 4 answers
Ownership
3.7 · Owned the fix and the follow-up test
Clarity
3.2 · Strong summaries; ran long on Q2
Flags
None

Example output for a sample candidate.

Developers

Build experiencesthat feel alive.

One API call gives your product a face, a voice and a memory. You own the logic; we handle rendering, perception and real-time media.

APIs
REST for setup, WebRTC for the live session, webhooks for results.
SDKs
TypeScript, Python and drop-in React components.
Bring your brain
Claude by default, or any LLM and tool set you already run.
Custom personas
Your face, voice, wardrobe and knowledge, approved by you.
Integrations
ATS, LMS, CRM and calendar connectors, plus S3 export.
Real-time infra
Edge-routed media with global turn latency targets.
1import { ManavOS } from "@manavos/sdk";23const hos = new ManavOS({ apiKey: process.env.HUMANOS_KEY });45const interview = await hos.conversations.create({6  persona: "aarohi",7  brain: { provider: "anthropic", model: "claude" },8  rubric: "rb_data_analyst_v3",9  memory: { candidateId: "cand_0412" },10  webhook: "https://ats.example.com/manavos",11});1213// Send the candidate a join link, or embed it14console.log(interview.joinUrl);

SDK shown as designed for the public beta.

Enterprise

Built for theinterviewsthat matter.

Security
Encryption in transit and at rest, SSO/SAML and role-based access. Built toward SOC 2 Type II.
Privacy
Candidate data is never used to train shared models. Retention windows you set, deletion on request.
Disclosure by design
Every persona says it's an AI at the start of every conversation. Not configurable.
Fairness
Same questions, same rubric for everyone, with adverse-impact reporting on outcomes.
Scale & reliability
Thousands of concurrent sessions, regional media routing and a public status page.
Custom deployment
Dedicated tenancy, data residency options and private networking for regulated teams.

Contact

Talk to asolutions specialist.

Tell us what you’re hiring or building for. A person on our team, not a bot, will reply within one business day.

Prefer email? hello@manavos.ai

We’ll only use this to reply to you.

ManavOS · Aarohi-1

You haven’t metan interviewerlike this.