Automated candidate evaluations with client-side AI proctoring
IntelliProctor — AI-Powered Recruitment Hub
Next.js 14 App Router workspace pairing browser-based MediaPipe FaceLandmarker singletons with secure server-side Firebase Admin access rules and Judge0 remote compiler webhooks.
Timeline
Sep – Dec 2025
Authentication
Enterprise RBAC
Code Run SLA
Judge0 sandbox
Tech focus
Project gallery
Visual highlights
A quick look at the states, flows, and artifacts that made this launch feel polished.
Project features
Crafted to feel human, performant, and scalable
IntelliProctor is an enterprise-grade recruiter console that streamlines candidate coding evaluations. It features MediaPipe client-side gaze tracking, noise detection, copy-paste guards, and a robust multi-language compiler integration validating remote Judge0 webhooks securely.
Enterprise Auth & RBAC
Firebase Authentication session cookies are managed server-side. Cascading access checks restrict dashboard tools based on organizational roles (ORG_OWNER, HIRING_MANAGER).
Client-Side AI Proctoring
GPU-accelerated MediaPipe FaceLandmarker singleton tracks facial gaze in the browser. Integrates noise monitoring and browser Visibility APIs to buffer and flag focus violations.
Integrated Coding IDE
Multi-tab editor with local progress caching across five languages. Code compiles inside Judge0 with HMAC SHA-256 signature verification validating callback webhooks securely.
Group-First Question Architect
Questions are organized into logical groups mapped to difficulty tiers. Deletions run inside transactional batches to clean up grading models and key dependencies with no ghost data.
Services & deliverables
End-to-end partnership
Strategy, engineering, and activation workstreams stayed tightly aligned so the launch felt intentional.
Service
Auth & Data Architecture
Implemented a Zero-client-write structure. The application makes no client-side Firestore requests, running all database transactions exclusively inside secure Next.js Server Actions.
- RBAC Middleware
- Server Action RPCs
- Data Security Threat Model
Service
Proctoring Engineering
Designed the MediaPipe gaze tracking WASM pipeline and local ViolationBuffer queue, flushing logs to Firebase in 30-second intervals to minimize database write traffic.
- gaze tracking task hook
- Violation buffering hook
- decibel audio checks
Service
IDE & Compiler Integration
Created the code runner system and webhook listener. Implemented timed cryptographic HMAC verification on webhook payloads to guarantee execution security.
- Code storage adapter
- HMAC webhook listener
- polling handler
Accomplishments
Business impact & program achievements
We measure success with data, ceremonies, and the teams who keep the experience running.
Data Isolation
Zero Client-Write
Executing all database updates strictly from server-side wrappers completely eliminates direct client access vulnerability vectors.
Callback integrity
HMAC Validation
SHA-256 verification and timing-safe checks secure webhook callback endpoints from spoofing attempts.
Throughput scaling
500 Invites/Batch
CSV candidate processing leverages Zod parsing models and chunked Firestore transactions to onboard large groups reliably.
Landmarks
Milestones that shaped the roadmap
From discovery to public launch we ran fixed-length pulses with clear artifacts at every turn.
Auth & Stage Pipeline
Configured Firebase session cookie rules and staged access layers.
AI Proctoring Core
Built client-side vision tracking handlers and local violation buffer queues.
IDE & Callback Webhooks
Configured Judge0 sandboxed compiling queues and timing-safe signature checks.
Admin HUDS & Bulk CSV
Shipped organizational admin panels, progress track HUDs, and bulk CSV validators.