Computer vision in production
Connecting visual processing with backend services, AWS, and the tools used to operate the system.
- Context
- Professional work at Creai
- My role
- Engineering across the whole system
- Scope
- AWS → ML → backend → backoffice
Professional work at Creai. This is a contribution overview; client data and private implementation details are not included.
The problem
A vision model's results need to reach the people and processes that use them. That requires video and event processing, maintained services and tools to inspect and operate the system.
What I contributed
- I worked on AWS infrastructure and backend services for event processing, OCR and video analytics.
- I took responsibility for ML and computer-vision components, as well as the internal backoffice frontend and dashboards.
- The work included deployment, maintenance and troubleshooting. At different stages, I also took on product definition and prioritization responsibilities.
What shipped
A production AI system connecting visual processing, backend services and internal tools. My contribution covered practically the full technical scope of the project, from infrastructure to the interface.
The value is in connecting the pieces.
My responsibility extended to both sides of the model: the infrastructure and data that feed it, and the applications that make its output useful. That whole-system perspective helps keep engineering decisions connected to operational needs.
- 01Video + images
- 02Vision + OCR
- 03Events + APIs
- 04Backoffice
Conceptual overview, not a customer environment or an exact production architecture.