I am a Ph.D. student in Computer Science at the University of Sao Paulo, where I also earned my Bachelor of Science in Computer Science. My research focuses on Machine Learning and Data Mining, with a particular interest in Time Series Classification and Extrinsic Regression. My Ph.D. work explores the balance between efficiency and effectiveness in time series classification. Alongside my academic pursuits, I enjoy competitive programming, where I continue to learn about new Algorithms and Data Structures. I also have a passion for teaching, and I have experience as a tutor for Competitive Programming, as well as a Teaching Assistant for Introduction to Computer Science, Calculus 2, and several Advanced Algorithms courses.
Implemented modernizations of the Device Management for Meta's network database through alignment with current industry standards and optimization for future scalability and reliability. Developed expertise in backend development and network infrastructure utilizing Python and C++.
Jan 2022 - Apr 2022.The popularization of mobile sensors and other technologies that collect data continuously over time increased the demand for machine-learning algorithms on time series. However, despite several machine learning and data mining techniques proposed in the field during the last decades, there are still unsurpassed challenges in the literature. One of the most common tasks in this domain is time series classification. Another, which has attracted significant attention recently, is extrinsic regression, which has the best results from adaptations of classification algorithms. In this scenario, it is necessary to explore various algorithms to obtain a model that performs well when a new dataset is acquired, in which ROCKET and HIVE-COTE 2.0 stand out. ROCKET is a computationally efficient method that uses convolutional kernels to perform accurately. HIVE-COTE 2.0 is a meta-ensemble with an extremely high training time that reaches the best accuracy rates described in the literature. Therefore, the balance chosen by the presented algorithms between effectiveness and efficiency becomes visible and necessary for large volumes of data. Thus, this research project aims to analyze and explore the trade-off between efficiency and effectiveness in time series classification and extrinsic regression tasks. To that end, we will develop and evaluate new algorithms to understand how to deal with this balance in various classifiers in the literature.
Feb 2023 - Dec 2027 (expected).Explore the Optimum-Path Forest (OPF) supervised classifier in the time series classification domain. The OPF is a similarity-based technique never explored in time series domains. OPF was tested on 128 datasets from the UCR Time Series Classification Archive, using Dynamic Time Warping (DTW) and Euclidean distances as metrics. OPF results in fair to middling classification error rates (statistically comparable to the Euclidean distance nearest neighbor classifier).
Sep 2022 - Dec 2022. GitHub RepositoryIntroduction to Computer Science I (2019/1), Calculus II (2019/2), Laboratory of Advanced Algorithms II (2021/2), Laboratory of Advanced Algorithms I (2022/1), and Advanced Algorithms and Applications (2022/2) courses. Provided instructional support to professors by creating and grading exams, projects, and exercises and led weekly exercise-solving and concept review classes. Tutored over 300 students in their first contact with programming, calculus techniques, and advanced computer science topics, resulting in improved skill sets and understanding.
Gave C++, algorithms, and data structures lessons focused on Competitive Programming Collegiate Contests and Olympiad in Informatics to about 100 undergraduate students annually. Some classes are available on YouTube, which has over 15000 views. Managed team formations, lesson schedules, problem sets, and local competitions for 4 years. Competed in several competitive programming contests, such as ICPC, Facebook HackerCup, and Google CodeJam. Obtained good results, including two ICPC World Finals qualifications in 2021 and 2022.
Feb 2018 - Current. Group websiteContemplated fields of image pre-processing, curvature and robust features calculation, besides curves reconstruction by means of previously extracted important points. Project supported by FAPESP.
Oct 2020 - Sep 2021. GitHub RepositoryArticles and references made for helping the students from Advanced Algorithms disciplines in University of Sao Paulo. Contains begginner-friendly competitive programming topics such general tips, python and C++ guides, greedy algorithms and more.
Notion link.Wrote blog posts about cryptography and security attacks on Medium platform, participated in Capture the Flag competitions and watched lessons about basics of InfoSec related topics, such as networking, web security, reverse engineering and social privacy.
Feb - Dec 2018. Group websiteWent to weekly meetings and lessons about Python, key algorithms, machine learning, data pre-processing techniques and frameworks.
Feb - Dec 2019. Group websiteProject made for multimedia classes using Python. Given an image, it returns the most similar images from the dataset, applying some concepts seen in multimedia classes, such as a dictionary of visual words and local feature extraction.
July 2021. GitHub RepositoryProject made in Python for Digital Image Processing classes. Aims to recognize a chess position using an image of a chess board as input, applying some techniques of Digital Image Processing and a Convolutional Neural Network.
May - July 2021. GitHub RepositoryProject made in C++ and GNUPLOT for Artificial Intelligence classes.
Sep - Oct 2020. GitHub RepositoryProject made in Java for Object-Oriented Programming classes.
May - June 2019. GitHub RepositoryTutorial written in Portuguese when I was a T.A. for Introduction to CS.
Apr 2019. GitHub RepositoryProject made for Evolutionary Systems classes using JavaScript that calculates the maximum possible value of a two variable function using different types of genetic algorithms.
Nov - Dec 2018. GitHub RepositoryProject developed in C, without using any auxiliary frameworks, for Multimedia classes. It compresses the original BMP file into a compressed binary .bin file, and can later decompress it back to BMP with a small perceptible difference to human eyes.
Apr - June 2021. GitHub RepositoryFinal project of my technical degree, aiming to influence the study of History in middle schools. Developed with Unity3D using C#.
Apr - Nov 2017.