DETECT - Deception Tracking Through Eye Cues Technology
Faculty-advised real-time deception detection platform using gaze and facial behavior signals.
DETECT was a 9-month, faculty-advised initiative where I led a 6-member team and contributed across research, infrastructure, and modeling.

Highlights
- Built a real-time AI-based deception detection system using MediaPipe FaceMesh and gaze analysis.
- Directed full-stack development with Astro (frontend), Go (backend), and PostgreSQL.
- Containerized with Docker and deployed on DigitalOcean VPS.
- Engineered gaze-tracking with Savitzky-Golay filtering and affine stabilization.
- Achieved 80%+ accuracy under the project’s evaluation setup.
- Managed Agile execution in Jira/GitHub, including sprint planning, review, and integration.
Technical Visuals



Resources
- Demo video: YouTube
- Frontend repo: bingKegeta/DETECT.js
- Backend repo: bingKegeta/DETECT.go
- Research repo: bingKegeta/DETECT
- Design document: DOCX
- Demo slides: PPTX