Loading...
AI Recruitment Automation
RecruitAI.
Enterprise-grade, AI-driven talent acquisition and secure assessment platform designed to automate the recruitment pipeline from application to final offer.
70%Time Saved
3x FasterScreening Velocity
90%Workflow Automation
50%Manual Labors Reduced
01 / Overview
Overview
RecruitAI is an enterprise-grade, AI-driven talent acquisition and secure assessment platform designed to automate the recruitment pipeline from application to final offer. The platform parses and scores candidate resumes using a multi-model LLM fallback mechanism, matching skills and experience directly to job descriptions. For secure examinations, it deploys a dual-camera proctoring system ("Third Eye") that streams desktop and mobile video feeds using WebRTC, utilizing Supabase Realtime for serverless signaling. Additionally, candidates undergo a voice-based AI interview round driven by browser-native speech-to-text and text-to-speech translation models. By combining automated screening, secure environments, and smart evaluations, RecruitAI establishes an efficient, bias-free, and highly scalable hiring workflow.
02 / Architecture Specs
Technical Specifications
| Component | Technology / Tool | Description |
|---|---|---|
| Framework & Client | Next.js 16 (App Router), React 19, TypeScript | Server-side rendering (SSR), client-side interactivity, and strict static typing. |
| Backend & Storage | Supabase (BaaS) | Database client, secure file storage (resumes/photos), and auth management. |
| Database | PostgreSQL (hosted on Neon Serverless) | Relational database housing applications, proctoring metrics, and exam sessions. |
| AI Assessment | OpenRouter (DeepSeek Chat, Qwen, Llama 3) | Multi-LLM fallback pool for ATS resume evaluation and automated interview grading. |
| WebRTC & Streaming | LiveKit & Browser WebRTC | Primary webcam streaming via LiveKit; side-angle phone camera via peer-to-peer WebRTC. |
| Signaling Protocol | Supabase Realtime Broadcast | Zero-cost serverless signaling handshake for peer-to-peer mobile camera feeds. |
| Browser Lockdowns | Safe Exam Browser (SEB) Configuration | Generates XML config profiles to lock down candidate systems during exams. |
| Speech Processing | Web Speech API (SpeechSynthesis, webkitSpeechRecognition) | Runs native Text-to-Speech (TTS) and Speech-to-Text (STT) during automated interviews. |
| OCR & Parsing | HuggingFace APIs, Mammoth, PDF-Parse | Parses text from resumes (.pdf, .docx) using cloud and local fallbacks. |
| Notifications | Nodemailer & SMTP | Sends automatic application results and exam invitations to candidates. |
03 / Platform Modules
Core Features
01
AI-Powered ATS & Resume Screener
- Zero-Failure File Parser: Supports PDF and DOCX parsing, routing requests through HuggingFace OCR with instant local pdf-parse and mammoth fallbacks if the primary API fails.
- LLM Evaluation Pool: Evaluates candidate resumes against standardized criteria (skills match: 40%, experience: 30%, education: 20%, clarity: 10%) using multiple LLM endpoints via OpenRouter.
- Automatic Shortlisting: Automatically transitions candidates to SHORTLISTED or REJECTED status based on their score threshold (70%) and triggers automatic email responses.
02
Dual-Camera Proctoring System ("Third Eye")
- LiveKit Integration: Establishes a primary video/audio stream from the candidate's desktop webcam directly to the recruiter's monitor desk.
- Mobile Camera Connection: Generates a temporary, token-secured QR code. When scanned by a candidate’s smartphone, it initiates a mobile web-camera stream showing a side view of their physical workspace.
- Serverless Signaling: Uses Supabase Realtime broadcast channels to exchange SDP offers, answers, and ICE candidates between desktop and mobile devices, avoiding the cost of a dedicated WebRTC signaling server.
03
Safe Exam Browser (SEB) Lockdowns
- System Locking: Generates and downloads a custom .seb plist XML profile that forces the candidate to take the exam in a locked sandbox, preventing screenshots, virtualization, Alt+Tab, and right-clicks.
- Real-time Fraud Checks: Monitors tab switches, copy-paste attempts, and microphone activity.
04
Interactive Voice-Based AI Interviewer
- Browser-Native TTS/STT: Asks questions using the browser's speechSynthesis and transcribes candidate answers in real-time using webkitSpeechRecognition.
- Automated Evaluation: Submits the text transcript logs to LLM models to analyze correctness and generate a final PASS/FAIL recommendation based on technical performance.
04 / Systems Flow
Architecture & System Flow
System Interaction Diagram
Table 01
profiles
Stores core candidate details, academic credentials, skills, and target job roles.
Table 02
applications
Manages the candidacy state (ats_score, ats_status, resume_url, status tracking).
Table 03
exam_proctoring_sessions
Tracks live integrity variables (e.g., mobile_connected, tab switches, microphone logs).
WebRTC Signaling Sequence
- 1Handshake InitializationThe Candidate Desktop page subscribes to a Supabase broadcast channel named
proctor-{examId}. - 2QR ScanThe candidate scans the QR code containing a JWT token. The phone loads the
/third-eye/connectroute. - 3Token ExchangeThe phone validates the token, subscribes to the
proctor-{examId}broadcast channel, and fires amobile-connectedmessage. - 4SDP ExchangeThe desktop receives the announcement, acts as the WebRTC receiver, and sends a connection
readymessage. The phone generates a WebRTC SDP Offer and transmits it via the channel. - 5Direct StreamThe desktop returns an SDP Answer and exchanges ICE Candidates via Supabase. Once connected, a secure, peer-to-peer MediaStream is established, piping the mobile camera feed directly to the Recruiter's Live Monitor client.