I was part of the AI ​​and Computer Vision R&D team, where I conducted several proofs of concept (PoCs) in areas such as suspended load detection and segmentation (which became a product), forklift detection and creating automatic red zones around these machines. I performed several experiments with autoencoder models, anomaly detectors, and computer vision techniques to improve our train track obstruction detection algorithm. I also participated in competitions, including one focused on identifying anomalies in ties and insulators in the electrical sector.Additionally, I started developing an automatic training pipeline by creating a data lake and using MLflow to record the experiments carried out. I was responsible for creating and supporting cloud asset protection operations, with an emphasis on detecting intrusions of people and vehicles, configuring VMs to run Deep Learning models with PyTorch and Ultralytics, in addition to using Pub/Sub, Cloud Functions and implement observability practices.

I was part of the AI ​​and Computer Vision R&D team, where I helped carry out several proofs of concept (PoC) on topics such as facial and gait recognition, ergonomics verification (which became a product), PPE (Personal Protective Equipment) verification and helmet color detection, in order to check whether a person is authorized to be in a certain location.I also worked on refactoring a computer vision pipeline to a filter architecture using Python, OpenCV, PyTorch, and Firebase.

Development and review of evaluation projects, in addition to conducting monitoring to clarify doubts about the content of the Pattern Recognition course.

As a Scholarship holder at the Voluntary Institutional Scientific Initiation Program (PIVIC), I am deeply engaged in pioneering research aimed at advancing noise reduction methods in low-dose computed tomography (CT) and accelerated magnetic resonance imaging (MRI). My role involves the implementation of cutting-edge deep neural network models to achieve noise reduction in these imaging modalities, with comprehensive evaluations conducted within esteemed databases like the Mayo Challenge Low Dose CT and FastMRI.

Scholarship holder of the Initiation Program in Technological Development and Innovation (PIBITI/PIVITI) working on research and development of a system to classify the body condition score of cows for a UFAPE zootechnical project to help local dairy farmers.

Scholarship holder of the Initiation Program in Technological Development and Innovation (PIBITI/PIVITI) working on the development of a web system for evaluating chicken eggs for the poultry department at UFAPE.

I actively contributed to the development and maintenance of robust APIs using Flask and FastAPI frameworks, utilizing MongoDB and PostgreSQL databases within a Domain-Driven Design (DDD) architecture. These APIs enhanced integration with the PontoTel system, empowering customers to streamline their operations effectively.Additionally, I played a key role in developing a machine learning classifier to categorize moments of the work journey based on the datetime of each moment and other relevant data from the worker. This involved preprocessing the data for optimal model performance and seamlessly integrating the machine learning model into the PontoTel product to enhance its functionality. The implementation of this model reduced manual work for the clients' HR teams by 80%, significantly decreasing the time spent on manually assigning journey moments.

I actively contributed to the development and maintenance of robust APIs using Flask and FastAPI frameworks, utilizing MongoDB and PostgreSQL databases within a Domain-Driven Design (DDD) architecture. These APIs enhanced integration with the PontoTel system, empowering customers to streamline their operations effectively.