Smart Security Patrol
SeeCure
With SeeCure, I worked on the full path from a simulated patrol environment to an analyzed incident and an observable cloud workload.
View source on GitHubContext
Problem, constraints, and my role
Security teams need earlier, more consistent visibility into incidents across autonomous patrol workflows.
Constraints
- The application connected simulation, vision analysis, and a distributed web platform.
- Detection quality and response latency both mattered.
- Cloud workloads needed centralized monitoring.
My role
I developed the distributed application and deployed its computer-vision workloads with centralized CloudWatch monitoring.
Project media
The system in view
Interface studies and working demonstrations from the project.
Architecture note
Patrol-to-observation flow
CARLA produces patrol input for YOLOv8 analysis; Django REST coordinates application data with PostgreSQL and Redis, while React presents the workflow and CloudWatch monitors deployed workloads.
Engineering decisions
The choices that shaped the work
Connect simulation to analysis
CARLA and YOLOv8 made it possible to exercise the patrol and detection workflow as a connected system.
Separate application concerns
Django REST, React, PostgreSQL, and Redis divide service, interface, persistence, and fast data access responsibilities.
Observe deployed workloads
Centralized CloudWatch monitoring kept the computer-vision workloads visible after deployment.
Result & reflection
What the work produced
Achieved 92%+ detection accuracy while reducing response latency by 30%.
What stayed with me
SeeCure reinforced that a real-time AI workflow is strongest when detection, response, and monitoring are considered together.
- Django REST
- React
- YOLOv8
- CARLA
- PostgreSQL
- Redis
- AWS