http://repositorio.ufc.br/handle/riufc/24009 Wed, 10 Jun 2026 19:40:41 GMT 2026-06-10T19:40:41Z http://repositorio.ufc.br/handle/riufc/86237 Título: Gerenciamento do ciclo de vida em LLMOPS: um survey sobre práticas e interações contínuas Autor(es): Sousa, Dirlia Vieira Abstract: Large Language Model Operations (LLMOps) emerges as an approach focused on managing the lifecycle of applications based on Large Language Models (LLMs), encompassing stages ranging from data preparation to deployment, monitoring, and the continuous evolution of models in production environments. Despite the recent growth of the field, the literature remains fragmented, with studies often addressing these stages in an isolated or linear manner, which hinders a comprehensive understanding of LLMOps as an integrated operational process. In this context, this work aims to analyze, through a literature survey, how the LLMOps lifecycle is addressed in existing studies, with a particular focus on practices and continuous interactions between its stages, as well as on its application across different contexts. The adopted methodology is based on systematic steps for searching, selecting, and analyzing primary studies, following predefined criteria to ensure methodological rigor and consistency in the synthesis of results. As a result, 13 studies were selected, from which it was possible to identify and analyze the main practices, processes, and continuous interactions that comprise the LLMOps lifecycle, as well as to understand its application across different domains. Tipo: TCC Thu, 01 Jan 2026 00:00:00 GMT http://repositorio.ufc.br/handle/riufc/86237 2026-01-01T00:00:00Z http://repositorio.ufc.br/handle/riufc/84734 Título: Avaliação do impacto computacional e energético do streaming bidirecional via gRPC no offloading multi-linguagem em ambiente de mobile cloud computing Autor(es): Farias, Miguel Barbosa Abstract: The increasing complexity of mobile applications imposes significant challenges on mobile devices, which are limited in computational and energy resources. In this context, computational offloading has been employed to transfer intensive tasks to remote servers, thereby reducing energy consumption and improving the performance of mobile devices. In parallel, recent advances in communication have enabled the emergence of streaming-based models capable of supporting continuous processing almost in real time. However, most solutions still rely on the traditional request–response model, without exploring the potential benefits of bidirectional streaming and multi-language strategies. Despite the potential of gRPC, its practical application in mobile and multi-language environments remains underexplored, particularly regarding its impact on performance and energy efficiency. This work proposed, implemented, and evaluated a multi-language computational offloading solution based on gRPC with bidirectional streaming, integrating an Android client in Java with servers implemented in Go and Java. The methodology involved applying video processing filters and comparing local and remote execution through metrics such as execution time, CPU usage, and energy consumption. The results demonstrated that the effectiveness of offloading directly depends on task complexity: for the simple filter (Grayscale), local execution proved more efficient, whereas for the more complex filter (Pencil), offloading provided significant gains. In the best evaluated scenario, using the Go server and the 5 GHz Wi-Fi network, the Pencil filter at 1080p resolution achieved a reduction of approximately 69% in execution time, accompanied by lower CPU usage and reduced battery consumption. It was concluded that offloading via gRPC with bidirectional streaming is both feasible and effective for computationally intensive mobile applications, provided it is applied to complex tasks under favorable connectivity conditions. The decision between local and remote execution should be adaptive, considering task complexity, content resolution, and network conditions, in order to maximize performance and energy efficiency in Mobile Cloud Computing environments. Tipo: TCC Thu, 01 Jan 2026 00:00:00 GMT http://repositorio.ufc.br/handle/riufc/84734 2026-01-01T00:00:00Z http://repositorio.ufc.br/handle/riufc/84639 Título: Análise e aprimoramento do modelo matemático para o problema de alocação de salas no contexto dos cursos de tecnologia da UFC — campus Crateús Autor(es): Furtado Neto, José Abstract: The Classroom Assignment Problem (CAP) consists of constructing a timetable that consistently assigns classes, professors, time slots, and rooms while satisfying a set of institutional and operational constraints. Due to its combinatorial nature, the CAP can be addressed through mathematical optimization models. In this context, this work aims to analyze the mathematical model proposed by Sousa (2019) for the CAP in the scope of the Federal University of Ceará — Crateús Campus, as well as to propose improvements that ensure greater structural consistency and practical feasibility of the obtained solutions. A detailed analysis of the original model revealed an inconsistency that allows a single class to be simultaneously assigned to the same day and time slot in different rooms, leading to impractical solutions. To address this issue, a new constraint is introduced into the mathematical model, ensuring that, for each (day, time slot) pair, a class is associated with at most one room. Computational experiments were conducted using the SCIP solver through the OR-Tools framework, considering different groups of instances and model configurations. The results demonstrate that the proposed constraint is effective in eliminating the identified inconsistency, thereby enhancing the structural coherence of the model. However, its inclusion increases the strictness of the problem, resulting in higher computational effort. Additionally, a sensitivity analysis was performed with respect to the professors’ preferred teaching days parameter. The results indicate that relaxing this constraint significantly improves the model’s resolvability, enabling the solver to conclude a larger number of instances either by finding optimal solutions or by certifying infeasibility. Overall, the study highlights the importance of balancing structural consistency and computational performance, and points to future research directions focused on scalability and deeper investigation of problem parameters. Tipo: TCC Thu, 01 Jan 2026 00:00:00 GMT http://repositorio.ufc.br/handle/riufc/84639 2026-01-01T00:00:00Z http://repositorio.ufc.br/handle/riufc/84630 Título: ARAMIS: uma ferramenta multiagente integrada com LLM Open-Source para apoio à correção de TCCs de estudantes de graduação ARAMIS: a Multi-agent Tool Integrated Autor(es): Sousa, Gustavo Campelo de Abstract: The correction of Undergraduate Final Projects is a crucial stage in the academic development of undergraduate students. However, this process can be time-consuming and exhausting both for students during their research activities and for advisors during supervision, due to factors such as task overload and insufficiently specific feedback on research content. The automation of scientific writing correction using techniques such as Machine Learning (ML) and Natural Language Processing (NLP) has become increasingly present in students’ daily routines, especially after the emergence of Large Language Models (LLMs). In this study, Academic Review Agents for Methodological Improvements ARAMIS was developed, a tool for the analysis and correction of undergraduate theses in portuguese, composed of three specialized agents: grammatical correction, logical chaining and methodological rigor, integrating an open-source LLM guided by prompt engineering techniques. A comparative analysis approach was adopted to evaluate the feedback generation by proprietary and open-source LLMs, with the objective of selecting a model that operates with a satisfactory trade-off. The proposed solution consisted of integrating the best-performing evaluated open-source model into ARAMIS, developed within the scope of this study, focused on returning the analyzed undergraduate theses feedback in portuguese, being composed by the three agents, which serve as the core pillars of automated revision generation. The tool receives the student’s text, which is processed by the LLM and returns a structured review based on the guidelines defined in the agent’s configurations. In this work, the System Usability Scale (SUS) was employed to assess usability. The experiments were conducted with real users actively engaged in undergraduate thesis writing, and the SUS questionnaire was applied immediately after tool usage. The results demonstrate that ARAMIS achieved a score of 90.5/100, confirming that the application meets undergraduate students’s usability expectations, being useful and achieving precise and targeted feedback. Tipo: TCC Thu, 01 Jan 2026 00:00:00 GMT http://repositorio.ufc.br/handle/riufc/84630 2026-01-01T00:00:00Z