• Evaluation of OCR (Optical Character Recognition) applications.• Fine-tuning of models such as llama, falcon and gpt.• Retrieval Augmented Generation (RAG) on a large corpus of text.• Research on image segmentation and large language models.• Implementation of face recognition/detection techniques• Handwritten recognition of old corpus of text

• Conducted A/B testing on different recommendation algorithms and techniques and authored a data quality checker using Great Expectations.• Contributed significantly to the success of the platform through highly praised work on the recommendation system of articles at Globo home.

• By utilizing advanced deep learning models, we successfully detected fraudulent modifications in driving license and identity card documents, resulting in an improvement of 30% in accuracy.• Bug fixes and software optimization through automation and integration, resulting in 20% time savings.• Highly praised for work on the deep learning-based fraud detection model, which has contributed significantly to the overall success of the company.• Successfully deployed multiple machine learning models into production environments.

Pix Force develops solutions using computer vision technologies, artificial intelligence, and machine learning. #1 Computer Vision and multi-award winning Startup in Brazil.• Collaboration with cross-functional teams to integrate deep learning models into existing systems and processes.• Development of models for image segmentation and classification tasks.• A significant improvement was accomplished in the OCR model. The model’s accuracy was boosted from an initial 70% to a remarkable 95%.

This project was supported by a research scholarship provided by the National Council for Scientific and Technological Development in partnership with the private sector and Federal University of Uberlândia.• Research in Deep Learning and computer vision applied to agricultural images, obtained by unmanned aerial vehicles (UAVs).• Exploitation of alignment methods for 5-channel images.• Implementation of style transfer method in order to align the images.• Publication of a high-impact paper at the end.

This Scientific Initiation project was supported by a research scholarship provided by the National Council for Scientific and Technological Development.• Optimization of data augmentation policies using Bayesian and random methods to train Convolutional Neural Networks (CNNs).• Image synthesis using Generative Adversarial Networks (GANs). • All the techniques were applied in the field of medical image analysis.

• Technical support for computer student users.