Skip to main content

AI-Conducted Interviews

Lira conducts interviews autonomously — deploying a bot to Google Meet, asking structured questions, evaluating candidates, and producing scored reports.

Interview Modes​

ModeDescription
SoloFully autonomous — Lira conducts the entire interview
CopilotCoaches a human interviewer with suggested questions and real-time feedback
ShadowObserves silently and generates a scored evaluation afterward

Lifecycle​

draft → scheduled → bot_deployed → in_progress → evaluating → completed

Create an Interview​

POST /lira/v1/orgs/:orgId/interviews
Authorization: Bearer <jwt>

{
  "title": "Senior Frontend Engineer",
  "mode": "solo",
  "scheduled_at": "2026-03-30T14:00:00Z",
  "meeting_url": "https://meet.google.com/abc-defg-hij",
  "job_description": "We're looking for a senior frontend engineer...",
  "skills": ["React", "TypeScript", "System Design"],
  "questions": [...]
}

AI Features​

Question Generation​

Generate tailored questions from a job description:

POST /lira/v1/orgs/:orgId/interviews/generate-questions

{"job_description": "...", "skills": ["React", "TypeScript"], "count": 8}

AI Draft​

Generate a complete interview configuration from natural language:

POST /lira/v1/orgs/:orgId/interviews/draft

{"prompt": "Create a technical interview for a senior backend engineer with Go and PostgreSQL experience"}

Resume Parsing​

Upload and parse a candidate's resume:

POST /lira/v1/orgs/:orgId/interviews/:id/resume
Content-Type: multipart/form-data

PDF → pdf-parse → GPT-4o-mini structured extraction → stored in S3.

Evaluation​

Phase 1 (Automatic)​

Runs automatically after the interview ends. Extracts Q&A summary from the transcript.

Phase 2 (Admin-triggered)​

Scores each criterion 0–100. Maximum 3 re-evaluations allowed.

POST /lira/v1/orgs/:orgId/interviews/:id/evaluate

Decision Recording​

PUT /lira/v1/orgs/:orgId/interviews/:id/decision

{"decision": "hire", "notes": "Strong technical skills, great culture fit"}

Decisions: hire, no_hire, next_round, undecided

Auto-Scheduler​

The backend polls DynamoDB every 30 seconds for interviews with scheduled_at due and auto-deploys the bot. No manual intervention needed.

Multi-Round​

Link successive interview rounds via parent_interview_id and round field:

GET /lira/v1/orgs/:orgId/interviews/:id/related