Research

  • Research Project — Undergraduate Research (2025)

    Benchmarking APIs across Different Programming Environments — A Financial System of Debits and Credits

    Final report (PDF, in Portuguese)

    Project data:

    Institution Federal Institute of Education, Science and Technology of Santa Catarina (IFSC) — Lages Campus
    Programme Undergraduate Research — Call no. 04/2024/PROPPI (rolling admission)
    Coordination Leonardo Bravo Estácio
    Student researchers Fabricio Iaronka de Souza and Vinicius Souza Corbellini
    Period January to December 2025
    Keywords Software Benchmarking; API Performance; RESTful APIs; Load Testing; System Scalability

    Summary:

    Implement a functionally equivalent RESTful API in several languages, following the same specification and the same architecture. Each variant ran on two instances, behind a load balancer using round-robin and over a shared PostgreSQL database, with identical CPU and memory limits, so that operating conditions stayed uniform. Load tests with a high volume of concurrent requests were run on that environment and repeated to ensure statistical validity, measuring latency, error rate, data consistency and CPU and memory usage. The goal is to produce empirical evidence to support the choice of languages and technologies when building scalable, efficient and resilient APIs.

    Results:

    Four equivalent implementations of the API were delivered (Java/GraalVM/Quarkus, JavaScript/Node/Fastify, Python/Uvicorn/FastAPI and Rust/Tokio/Axum), running on two instances balanced by Nginx over the same relational database. The test campaigns, run with Gatling in warm-up, steady-state and cool-down phases, produced series of latency (p50, p95 and p99), throughput, error rate and resource usage, analysed with non-parametric tests and bootstrap confidence intervals. The results show that persistent connections, prepared statements and a well-tuned connection pool reduce tail latency and variability under pressure, and that the transport format and serialisation can shift p95 and p99 substantially even with the database and the mix of operations held constant.

    Dissemination:

    The work was presented as a video and a poster at the 7th “3 Minutes of Science and Technology” Showcase, part of the 22nd National Science and Technology Week, held by IFSC (Lages and Urupema campuses), on 4 November 2025.

  • Research Project — Technological Development and Innovation (2025)

    Development and Evaluation of an Educational Application for Teaching Discrete Mathematics in Higher Education

    Final report (PDF, in Portuguese) Paper (PDF, in Portuguese)

    Project data:

    Institution Federal Institute of Education, Science and Technology of Santa Catarina (IFSC) — Lages Campus
    Programme Technological Development and Innovation — Call no. 04/2024/PROPPI (rolling admission)
    Coordination Leonardo Bravo Estácio
    Student researchers Caio Picinini Pagliosa and Eduardo Moreira de Jesus
    Period 2025
    Keywords Discrete Mathematics; Educational Application; Higher Education; Mobile Learning; Usability Evaluation

    Summary:

    Develop and evaluate an educational application for teaching Discrete Mathematics in higher education. The core problem is the weak consolidation of abstract concepts and the lack of timely feedback on exercises. The proposal covers the design and implementation of a cross-platform prototype, the evaluation of usability and quality in use, and the assessment of its impact on academic performance through pre- and post-tests with independent groups. The method organises successive cycles of building and evaluating, with usability testing and learning measurement, aiming at a validated artefact, replicable design guidelines and evidence of learning gains.

    Results:

    The project delivered DiscMath, a cross-platform prototype built in Flutter and organised into content and practice modules, with interactive questions, scored mock exams, support material by topic, immediate feedback and engagement features such as rooms and rankings. Usability was assessed with the System Usability Scale (SUS) among higher education students, reaching an average score of 73.75 — above the adopted acceptance threshold (SUS ≥ 68) and rated as satisfactory. The quasi-experimental study of the impact on academic performance, planned in the project, was not completed within the cycle and remains as a next step.

    Dissemination:

    The work “Developing DiscMath: an Application for Teaching Discrete Mathematics in Higher Education”, by Caio Picinini Pagliosa, Eduardo Moreira de Jesus, Vilma Gisele Karsburg and Leonardo Bravo Estácio, was presented as a video and a poster at the 7th “3 Minutes of Science and Technology” Showcase, part of the 22nd National Science and Technology Week, held by IFSC (Lages and Urupema campuses), on 4 November 2025.

  • Master's Dissertation (2014)

    On the Choice of Relaxation and Ordering of the Projections in Kaczmarz's Method with Emphasis on Highly Parallel Implementations and Applications in Tomographic Reconstruction

    Dissertation (PDF, in Portuguese)

    Project data:

    Institution Institute of Mathematics and Computer Sciences (ICMC), University of São Paulo (USP) — São Carlos/SP
    Programme Computer Science and Computational Mathematics
    Supervisor Prof. Dr. Elias Salomão Helou Neto
    Defence June 2014 (72 pages)
    Funding Scholarship from the São Paulo Research Foundation (FAPESP)
    Keywords Kaczmarz; Relaxation Parameter; Numerical Methods; Parallel Algorithms

    Abstract:

    The Kaczmarz method is an iterative algorithm for finding the solution of a system of linear equations Ax = b by projecting onto the hyperplanes, widely used in applications involving Computerized Tomography. It has been recently highlighted after the publication of a random version presented by Strohmer and Vershynin in 2009 that yields probably exponential convergence in expectation. Thereafter, Eldar and Needell in 2011 suggested a modified version of Strohmer and Vershynin algorithm, which at each iteration selects the optimal projection from a random set making use of the Johnson-Lindenstrauss lemma. None of the mentioned articles presents a technique for choosing the relaxation parameter, however, the proper selection of this parameter can achieve a substantial gain on the speed of the method. In this project we present a methodology for finding the relaxation parameter, as well as parallel implementations of Kaczmarz’s Algorithm using the ideas of Eldar and Needell. Our methodology for parameter selection uses a new generalization on Strohmer and Vershynin’s results which now regards the relaxation parameter λ. Thenceforward, we obtain an estimate of the convergence rate as a function of λ. Then we use this estimate in the algorithm the optimizer of this estimate. The parallelization of the methods has been implemented through the CUDA platform and appears to be very promising, since it delivers substantial gain in the convergence speed.