2026-08-132026-08-130026-07-17BEZERRA, Romulo Lobato. Desenvolvimento de uma plataforma web assistida por inteligência artificial para o ensino da língua inglesa. Orientador: Iago Lins de Medeiros. 2026. [16] f. Trabalho de Curso (Bacharelado em Engenharia da Computação) – Faculdade de Engenharia da Computação, Campus Universitário de Tucuruí, Universidade Federal do Pará, Tucuruí, 2026. Disponível em: https://bdm.ufpa.br/handle/prefix/9921. Acesso em:.https://bdm.ufpa.br/handle/prefix/9921This work presents the development, architecture, and validation of the ChatEP platform, an open-source web ecosystem designed for autonomous English learning. Based on the Task-Based Language Teaching (TBLT) peda gogical approach, the system orchestrates heterogeneous specialized Artificial Intelligence models (Flan-T5 Large, Whisper, and TinyLlama-1.1B) fine-tuned via Low-Rank Adaptation (LoRA), aiming to improve listening, writing, spea king, and reading skills. Quantitative validation was based on data telemetry metrics, applying the BLEU index for textual correction, the WER rate for the acoustic pipeline, and the BERTScore to evaluate the virtual tutor’s semantic si milarity. Additionally, a pilot study with real users validated usability and map ped architectural trade-offs regarding synchronous cloud inference latency. The experimental results obtained demonstrate high levels of accuracy and techni cal compliance of the ecosystem against the specified requirements and business rules.Acesso AbertoEngenharia de softwareMicrosserviçosFine-tuningLow rank adaptationProcessamento de linguagem naturalHomologação de sistemasValidação de arquiteturaSoftware engineeringMicroservicesLow-rank adaptationNatural language processingSystems validationArchitectural trade-offsCNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::METODOLOGIA E TECNICAS DA COMPUTACAO::ENGENHARIA DE SOFTWARETrabalho de Curso - Graduação - ArtigoAttribution-NonCommercial-NoDerivs 3.0 Brazil