João Antônio Marcolan Email & Phone Number

Porto, Portugal

I was a part of the Scanner project, which involved processing and analyzing documents to extract specific elements. During this project, we encountered high error rates in the detection, and the processing capacity was limited, affecting both performance and client satisfaction.Task:The task was to develop and propose new features that would improve tag detection, increase the processing capacity of PDFs, and increase the reliability of the whole system.Action:To achieve these goals, I took the following actions:1. Feature Development: I suggested and created new features for the Tag Scanner project that focused on error reduction and improved tag discovery. This included the development of visualization and debugging tools, which helped in identifying and resolving issues more efficiently.2. Library Creation: I extracted the most important parts of the program to create a new library that adhered to best practices, ensuring modularity and reusability.3. Performance Optimization: I implemented innovative solutions to enhance performance, resulting in a 10-fold increase in the capacity to process PDFs. This also improved the project's reliability and reduced the error rate.Result:The Tag Scanner project saw significant improvements as a result of these actions. In the last evaluation, the error rate in tag detection dropped from 19.84% to just 1.89%, while the project identified 25.39% more tags, and the capacity for processing PDFs increased tenfold. The programs became more reliable, reducing the number of reprocessing requests, which led to cost reductions in human and infrastructure resources.

Brazil

I worked on a project for a contracting enterprise where the goal was to develop a product using image processing, computer vision, and optical character recognition (OCR) to extract information from engineering schematics. This product was designed to scan images of complex engineering schematics and retrieve structured information from them.Task:The main challenge was to develop a software solution that could accurately process engineering schematics and convert them into structured data. This required training machine learning models aligned with image processing techniques and traditional algorithms to create a robust and efficient system.Action:I was responsible for several key tasks:1. Integrated the machine learning model with image descriptors and traditional algorithms to create a cohesive software solution.2. Developed image descriptors to facilitate the extraction of meaningful information from engineering schematics.3. Implemented classical algorithms to use the machine learning model and the image descriptors to structure the data.Result:The final product achieved an accuracy rate of 75.15% in extracting structured information from engineering schematics. Additionally, the most important algorithms from this project were extracted to create a library. This library is now used to create new functionalities for other programs and improve debugging tools. This result was considered a significant success.

Florianópolis, Santa Catarina, Brazil

In the realm of behavioral studies, hypothesis testing has long been a popular method for predicting the effectiveness of treatments. However, this methodology has faced criticism for its limitations and misinterpretation.I was required to address these issues to ensure a more rigorous and safe use of hypothesis testing in animal behavior studies.Task:The primary objective was to improve the rigor and safety of using hypothesis testing in behavioral studies, given its widespread use and associated risks. It was necessary to develop new software tools, statistical models, and methodologies that would provide clearer insights and facilitate better decision-making. Action:1. Ecosystem Development: I designed EthoWatcher OS to be part of a broader ecosystem of programs, enabling interoperability with other software tools. This allowed researchers to conduct a wider range of analyses and apply different methodologies to the same data set.2. Machine Learning and Statistical Models: I constructed and validated various machine learning and statistical models to enhance the accuracy of behavioral analysis. This included exploring alternatives to hypothesis testing and integrating these models into the ecosystem.3. Concordance Analysis: I developed a methodology to calculate concordance and implemented it in software. This tool can help researchers identify discrepancies and improve the consistency of their results.Result:My efforts could lead to significant advancements in the rigor and reliability of animal behavior data analysis. The improvements in the EthoWatcher OS provided researchers with a more robust platform for managing and analyzing data. By acknowledging the limitations of hypothesis testing while offering safer usage guidelines and alternative methodologies, I helped increase awareness of the associated risks. The concordance analysis tool can be used to evaluate different types of classifiers and help refine the category catalog.

Florianópolis, Santa Catarina, Brazil

The development of EthoWatcher OS, an open-source program aimed at traditional behavioral analysis. This software allowed for traditional behavioral tracking and facilitated the extraction of kinematic and morphological descriptors from video data, enhancing the range of analyses that could be done with this software. Performing stimulus as quickly as possible was crucial for some behavioral studies, so the software needs to be capable of delivering a result as quickly as possible.Task:My project was to make a program that could handle video data in real time and have unique features, such as concordance analysis and sequential analysis. My job was to develop the software and make sure it met these requirements.Action:I focused on the development of EthoWatcher OS:1. Image Processing and Analysis: I implemented image processing techniques to extract kinematic and morphological descriptors from the video data. Additionally, I developed algorithms to speed up the data acquisition process.2. Parallel Programming: To achieve real-time performance for video recording and image processing, I employed parallel programming techniques to multithread different parts of the program.3. Innovative Algorithms: I proposed and developed unique algorithms not found in similar programs, such as double blinding, sequential analysis, and concordance analysis.Result:The resulting EthoWatcher OS program provided a comprehensive set of tools for behavioral analysis, including the ability to extract and analyze data not available in similar programs. The software's unique approach to data organization allowed researchers to perform concordance and sequential analyses with ease. This software propelled me to a PhD and resulted in some congress publications.

Doutorado, Bioengineering And Biomedical Engineering

Topic: Increased rigor in acquiring animal behavior data and models used in decision-making about the effect of medications on their.

Master Of Engineering - Meng, Biomedical/Medical Engineering

Open source tool for behavior analysis and real-time video acquisition using computer vision and parallel processing techniques Adivisor.

Engenheiro Eletricista Com Enfase Em Eletrônica.

Graduated First Class Honours Topic: Prototype of a remote mechanical arm controller (C, Matlab). Adivisor: Amauri Fagundes Balotin