This post marks the beginning of this research log. The goal of this blog is fairly simple: to serve as a public record of my PhD work in Computer Science, including ideas, experiments, questions, and lessons learned along the way.

I am currently a PhD student based in Campinas, Brazil, conducting my research at the University of Campinas (UNICAMP). My work focuses primarily on weak supervision for semantic segmentation, with an emphasis on understanding how dense prediction models can be trained under limited, noisy, or indirect supervision. Rather than assuming access to large, fully annotated datasets, I am interested in methods that better reflect the realities of applied research.

A significant part of my background is tied to seismic data analysis, particularly seismic facies segmentation. Working with seismic data has strongly influenced how I approach machine learning problems: labels are expensive and ambiguous, data distributions are messy, and performance metrics rarely tell the full story. This domain continues to act as both a motivating application and a stress test for the methods I study.

Alongside my research, I have a strong interest in the process of doing research itself. This includes experiment design, reproducibility, tooling, and building systems that make large-scale experimentation tractable. As part of this, I created and currently lead the development of an open research framework used within my academic environment to support structured and reproducible machine learning workflows. Many of the posts here will inevitably reflect this systems-oriented perspective.

In this blog, I plan to write about:

• Papers I am reading or have read, including informal summaries and critical reflections
• Ongoing research ideas, partial results, and open questions
• Technical details from experiments that may not make it into publications
• Notes on methodology, tooling, and research workflows

The posting frequency here may be somewhat sparse, especially during intense research or writing periods. That said, I value consistency over volume, and I intend to keep this space active as a long-term research diary rather than a high-frequency news feed.

If you are a student, researcher, or practitioner interested in semantic segmentation, weak supervision, or applied machine learning research, I hope some of these notes prove useful — or at least relatable. This blog is as much for thinking in public as it is for sharing finished ideas.

I am always open to discussion, collaboration, or informal exchanges around research topics covered here. You can find me on LinkedIn, explore my work on GitHub, or reach out directly via email.