Portfolio

Jusbrasil

At Jusbrasil, I played a key role in maintaining and evolving the company’s data platform. Our team undertook a complete refactoring of the platform, transforming it into a cloud data-lake capable of processing terabytes of data monthly. The project used then cutting-edge technologies including Kafka, BigQuery, Apache Spark, and a suite of custom-built tools.

The refactored platform significantly enhanced Jusbrasil’s data processing capabilities, enabling more efficient extraction, loading, and transformation of large-scale datasets. This improvement directly contributed to the company’s ability to handle and analyze vast amounts of legal information, supporting its mission to make legal processes more accessible and transparent.

In addition to my core data engineering responsibilities, I gained valuable experience in DevOps practices. This included operating a Kubernetes cluster, managing infrastructure through Terraform, and implementing CI/CD pipelines. These skills were crucial in ensuring the reliability and scalability of our data platform.

UVCS

UVCS is a project in development aiming to map and model pathogens under UV-C light to combat hospital infections. This map is continuously updated, informing our robot TALOS, how much time it should devote to each location to ensure a safe and thorough sanitation. It also aims to develop a web app where hospital managers can track how sanitized each point of each environment is, how many times has each space been sanitized and how much time is needed to clean it. The project won Best Team in the 2020/2021 Entrepreneurial Initiation Program selected by UFRGS and the Chamber of Industry (Senai RS).

I was the main developer of the embedded model, which was based on the works of Walker and Ko, Sabino et al and many others. The implementation was initially done in Python but later we switched to Julia for increased performance. In the system’s current iteration, we estimate a 99% reduction of the pathogens in the room in under one minute.

Pampa Aerodesign

Pampa Propeller Data Set

Pampa Propeller Data Set or PPDS, provides dynamic performance data for all propellers made by APC, in a analysis-friendly format, with propeller and RPM indexation. It enables the continuous update of this data from the latest files provided by APC via a Python Crawler and regex formater.

Features in development

A Streamlit app to visualize data from the propeller database;
Incorporating unit conversion with Pint;
Local indexing of the dataset using TinyDB’s JSON database;

Analysis

I have also analysed the propellers the team currently owns, using Pandas and Matplotlib to visualise their performance.

Airfoil Optimization

The team had the task of improving the performance of it’s airfoil, to do so we employed a genetic algorithm written in Fortran aiming to maximize the lift coeficient (CL) whilst maintaining a minimum thickness and a maximum drag coefficient. The computational costs of these simulations was too high and so I employed Google Cloud’s elastic computation service, thereby reducing the execution time by a factor of 10. I subsequently also analyzed the performance of the resulting airfoil, which had a 10% higher CL while holding to the stated optimization constraints.