At Instituto de Ciência e Tecnologia Itaú - ICTi, I developed and maintained machine learning workflows on AWS, using Amazon S3 and Amazon SageMaker for ETL, training, and inference.
I built and maintained a 700+ GB corpus from financial events and multiple data sources. I also developed an end-to-end ETL pipeline for tokenizing structured financial event data, including encoding and decoding.
I trained and ran inference with GPT models, evaluating them with loss per token and perplexity. My work also included time series monitoring and anomaly detection across products and financial events.
My research at Universidade do Estado do Rio de Janeiro focused on numerical and spectral methods for physics problems. I have an M.Sc. in Physics and have used Python for data modelling and visualization.
