Mejorando la seguridad y la fiabilidad de los procesos con reconocimiento facialImproving the security and reliability of processes with facial recognition

Staff attendance is a critical operation, prone to large financial losses from bad practices if it is not controlled. This was the first project as a Software Engineer that I led development and shipment of the final product, starting from facial-recognition SDK research with the mobile team. The solution has been adopted by more than 6,000 employees and backs more than five critical processes in the company, used for material management, fuel loads, material control and other flows.

The app had to work offline, run staff enrollment and facial recognition, and transmit that information under high security standards with liveness checks and regulatory compliance. I integrated Jetpack Compose and CameraX, used location and gyroscope sensors to guide enrollment and the user experience, and implemented computer-vision models with TensorFlow Lite for object detection.

The main technical challenge was running these demanding models on devices with limited resources. After the optimizations, registration time on low-to-mid-range hardware dropped from 14 seconds to about 7 seconds on average, and mid-to-high-range devices reached 3 to 4 seconds. I also packaged these processing modules for cross-platform apps, so they could serve as a digital signature across operational flows and substantially strengthen reliability and traceability.