Professional Experience
Instituto de Ciência e Tecnologia do Itaú
I work in the Itaú Unibanco’s innovation lab, where I am responsible for developing and implementing advanced data science solutions to drive innovation and improve customer experiences. I have developed Generative AI agents that help thousands of custumers with their payments. Used technilogies include Python, LangChain and LangGraph.
C6 Bank
I developed tools for the C6 chatbot, helping the business team to decrease the time needed to add new features. I also worked in a Vision project that was made available for other analysts to use.
Banco do Brasil
I was hired through Capgemini to work for Banco do Brasil (Bank of Brazil) as NLP Specialist. My main responsibilities included developing and implementing natural language processing models to improve internal processes and tools.
BigData
I worked on a recommendation project for the largest beer manufacturer in the world providing daily recommendations to more than 1million POS (point of sale). I maintained both the ML model, including techniques and training, and the infrastructure for serving the recommendations.
NeuralMind
During my time at NeuralMind, I worked on various deep learning projects, including vision, NLP and classic ML. I also participated in a few articles and won the RadVid19 challenge, which was a competition to diagnose COVID-19 using AI.
Papers
Posts
Awards
1st Place - RadVid19 Challenge
First place on all three proposed challenges: Classification of X Ray images as covid or non-covid; Classification of Computed Tomography as covid or non-covid; Segmentation of lung and opacities in Computed Tomography.
SIPAIM Society Award
Award related to the ‘Multitasking Segmentation of Lung and COVID-19 Findings in CT Scans Using Modified EfficientDet, UNet and MobileNetV3 Models’ paper.
8th Place - SROIE 2019
We achieved 8th place on the tasks of Text Localization and Scanned Receipt OCR on the Robust Reading Competition SROIE. Using a Unet and morphological operators we were able to localize the text on a scanned receipt. The detected text was then fed to a CRNN with CTC network to be recognized.
