Talent Systems — Science Team
Platform Overview

What's Built

Complete inventory of platform features as of the current sprint.

Last updated: 2026-04-14

Complete Features

Core Interview Platform

  • Structured (spine) interviews — voice and text modes
  • Adaptive interviews — AI generates questions from competency anchors
  • Interview depth: light (4), standard (6), deep (8), custom
  • Idle auto-end for voice (configurable 5s warning, 10s end)
  • Camera + face detection gate before interview starts
  • Resume upload with text extraction (PDF, DOC, DOCX, TXT, RTF)
  • Single-use, 7-day interview tokens
  • Mobile blocked for voice, allowed for text

Competency Framework

  • 8 seeded behavioral competencies with 5-level anchors
  • Database-managed (no hardcoded competencies)
  • Categories: behavioral, situational, technical, post-hire
  • Employer-created custom competencies (manual + AI-generated)
  • Competency profiles (pre-built bundles)
  • Admin + employer CRUD via shared UI components

Scoring System

  • Claude-based post-interview scoring (Haiku dev, Sonnet prod)
  • Per-competency scores (1-5) with evidence quotes, strengths, concerns
  • Weighted overall score based on posting competency weights
  • Recommendation: Advance / Hold / Decline (configurable thresholds)
  • Candidate name scrubbed before scoring
  • Scorer version stamped on every scorecard

Rubric System

  • Interviewer rubric (structured): 15+ configurable fields
  • Interviewer rubric (adaptive): 25+ configurable fields
  • Scorer rubric: 10+ configurable fields including 6 methodology fields
  • Integrity rubric: configurable weights and thresholds
  • AI generation rubric: configurable system/user prompts for JD, competency, question generation
  • All stored as JSON in rubrics table, editable via admin UI

Rubric Versioning

  • Semver: Revision (major), Refinement (minor), Calibration (patch)
  • Immutable version history with rollback
  • Version stamped on transcripts (interviewer_version) and scorecards (scorer_version + interviewer_version)
  • Trends: average scores, completion rates, recommendation breakdown per version
  • SearchableSelect version history browser

Integrity Monitor

  • 7 flag types: multiple_faces, extended_gaze_away, tab_hidden, window_blur, clipboard_paste, window_resize, multiple_displays
  • Configurable deduction weights per flag
  • Configurable gaze angle/duration thresholds
  • Browser-only processing — no biometric data transmitted
  • Admin configuration page with sliders

Employer Features

  • Multi-user accounts with 4 roles (owner, admin, member, associate)
  • Configurable per-role permissions (9 areas x 3 access levels)
  • Invitation-based onboarding (users set own passwords)
  • Team management: invite, remove, change role, transfer ownership
  • Dashboard with posting list, search, filters, sort
  • Draft postings (save incomplete, publish when ready)
  • Posting detail: pipeline funnel, analytics, heatmap, candidate list
  • Candidate comparison (2-5 side by side)
  • CSV + PDF export
  • Talent pool with vector relevancy scoring
  • Custom competencies and profiles

Admin Features

  • Platform-wide dashboard with stats
  • All postings across employers with filters
  • Employer management (create, suspend, reactivate)
  • Full rubric configuration (interviewer, scorer, adaptive, integrity, interview, AI generation)
  • Competency and profile management
  • Talent pool sync and relevancy settings
  • Analytics (per-employer usage tracking)
  • Flagged interviews
  • Audit log
  • User management (invite, deactivate admins)

Tests

  • 1,193 unit/integration tests across 86 files
  • ~130 Playwright E2E tests across 21 spec files
  • 86% lib statement coverage

On this page