Our work / Project Code
Project Code
Technical assessment and proctoring platform
Runs online technical assessments end to end: AI-generated questions and test cases, multi-layered proctoring during the exam, and automated scoring afterwards. Deployed on an on-premises OpenStack cloud.
- Category
- Hiring / technical assessment
- Platform
- Web
- Stack
- ReactNode.jsPostgreSQL
- AI services
- Vertex AIMediapipeCOCO-SSDVoice Activity Detection
- Infrastructure
- GitLab CI/CDDockeron-premises OpenStack
What the platform does
Capabilities
Custom assessments made simple
Design assessments around your own hiring needs — multiple choice, coding challenges and subjective questions on one intuitive platform.
AI-powered question generation
Generative AI suggests questions and produces coding test cases automatically, saving recruiter time while keeping quality and relevance.
Advanced cheat detection
Facial landmark detection observes candidate focus, eye movement tracking flags suspicious behavior, and background audio analysis catches unauthorized noise or conversation.
Instant scoring and analytics
Coding answers are analyzed for functionality against AI-generated test cases; MCQs and subjective answers score against predefined criteria.
Easy integration
A robust API framework connects Project Code to existing HR and recruitment systems for results, reports and data sync.
Role-based access
Managers contribute questions, reviewers approve content, administrators oversee assessments — accountability and security by design.
Modular, so each piece scales on its own
Architecture
- Frontend
- React — responsive interfaces for candidates and recruiters.
- Backend
- Node.js.
- Database
- PostgreSQL.
- Facial landmark detection
- Mediapipe tracks eyes, nose and mouth to monitor focus and flag unusual behavior.
- Object detection
- COCO-SSD identifies electronic devices or unauthorized materials in the candidate's environment, in real time.
- Test case generation
- Vertex AI creates diverse scenarios to evaluate a candidate's coding solution thoroughly.
- Audio monitoring
- Voice Activity Detection flags prolonged or unauthorized voice activity during an exam.
- Question distribution
- Randomized algorithms generate a unique set per candidate, reducing collusion and question sharing.
Two things a proctoring platform has to get right
Quality assurance, infrastructure and privacy
Validation
- Every AI-generated question and answer is manually validated by QA — generation is not assumed correct.
- QA works with developers and the AI research team on scenarios that simulate real candidate behavior.
- End-to-end UI testing with Cypress across both candidate and recruiter interfaces.
Infrastructure and data
- GitLab CI/CD for consistent build-test-deploy cycles, with apps containerized in Docker.
- Hosted on an on-premises OpenStack cloud, supporting local and in-house deployments.
- TLS/HTTPS in transit, role-based access control internally, plus regular backups, audits and monitoring.
Project Code does not currently claim formal compliance with frameworks such as GDPR or HIPAA. It is built around their core principles — transparency, accountability and data minimization — and we would rather say that plainly than imply a certification we do not hold.
Built in the same workshop
