Hybrid System for Artificial Intelligence Model Inference in Local and Cloud Environments, Oriented to the Optimization of Research Processes through Flexible and Innovative Infrastructure

This project proposes a hybrid architecture for artificial intelligence (AI) model inference, integrating the local infrastructure of the National Telecommunications Corporation (CNT) with scalable AWS services. The initiative aims to address critical challenges regarding scalability, latency, and operational costs associated with deploying large language models (LLMs). By combining local resources with the cloud, the system enables flexible workload management, ensuring that low-latency tasks are processed locally while demand spikes are handled in the cloud. The collaboration between the Salesian Polytechnic University (UPS) and CNT facilitates knowledge transfer and the practical application of advanced technologies. The project focuses on configuring a robust infrastructure, automated model synchronization, and the implementation of advanced monitoring systems such as AWS CloudWatch and Prometheus. Expected outcomes include significant improvements in response times, optimization of computational resources, and the creation of new service capabilities for end-users. Beyond technical benefits, the project promotes applied research and strengthens the professional profile of students through their direct participation in developing innovative solutions. The sustainability of the system is ensured through continuous resource optimization and predictive maintenance, establishing an efficient model for AI adoption in corporate environments.<br/><br/><b>Goal</b>: <br/>Develop and implement a hybrid system for artificial intelligence model inference that combines local infrastructure with AWS cloud services. The goal is to optimize performance, scalability, and operational efficiency in data processing.<br/><br/><b>Research lines</b>: <br/>Artificial intelligence and data mining