I write about technology, personal projects, entrepreneurship, and machine learning.
About me: Born in Sao Carlos, Brazil, I studied Computer Science at USP and earned a master's degree in Machine Learning at UFSCar. Throughout my career, I was partner-director and CTO at Raccoon, later CTO at Monks, and I am now founder and COO at Draiven. I also work as a technology and management advisor.
Article
For many years, the ERP was treated as the company’s central system. It was where teams registered orders, purchases, invoices, inventory, payments, receivables, production, contracts, master data, and a large part of the business’s critical operations. For many companies, implementing an ERP meant professionalizing management, standardizing processes, and gaining control. This is still true, but the role of the ERP is changing.
Article
Process Engineering has evolved from a factory-floor discipline into the essential architecture of the modern AI enterprise. While AI tools boost individual speed, these gains often fail to scale because faster employees don’t fix broken organizational flows.
Article
Most graph clustering methods fail because they ignore the network's shape, focusing only on node data. DGCSD solves this by using structural "seeds" to guide a Graph Neural Network, marking the first time seed detection and GNNs have been combined for clustering. By integrating both connection patterns and node features across its entire pipeline, the model creates sharper, more accurate groups. It is a competitive, open-source approach that proves understanding how nodes are connected is just as vital as knowing what they are.
Project
In this study, we propose DGCSS, a novel approach designed to push the boundaries of state-of-the-art performance in node clustering within complex networks. Traditional deep clustering methods often rely on external clustering algorithms to identify representative elements, but these methods can be hampered by the influence of initially formed groups and typically focus only on the content information of each example, neglecting the rich topological structure of the data.
Project
A Telegram-based personal assistant project, built from scratch for total control over LLMs, token consumption, and security, integrating Notion, Google Workspace, and health metrics into a single interface.