Vehicular congestion derives from the obstruction of the passage for cars circulating or moving on the roads. In the last five years, Lima has seen an increase in traffic and congestion. One of the solutions to this problem is the improvement of traffic light systems, but in Peru, there have been no studies that address these proposals. For this reason, this paper proposes the design of an intelligent traffic light system using computer vision and artificial intelligence concepts to reduce traffic congestion and its consequences at intersections in the streets of Lima. For the development of the article, an analysis of solutions, technologies and methods used to reduce congestion was carried out. In addition, the design of our proposed solution was carried out, presenting three architecture designs: architecture of components that would make up the system, architecture of integration with Cloud services for our system and an architecture of system layers. Also, the system of detection and time allocation for traffic lights is presented, validating its operation in a test environment, obtaining satisfactory results, such as 85.90% for accuracy, 59.40% for recall and 64.80% of mAP in the detection of the model and a time allocation related to the vehicular flow for the streets that were studied.