Machine Learning engineer on the MLOps team at Banco Votorantim, a client of the company ACT Digital. I work on the development of the Kraken platform, which aims to provide a controlled environment for the models submitted by data scientists, ensuring the correct functioning, data quality and maintenance of the proposed models. Within Kraken, develop various applications and improvements for the platform.Development of an automation for Kraken's CI pipeline, which analyzes the data scientist's code, extracts the docstrings that describe each function and class, collects relevant information about the metadata behind the code, processes the file so that this This step is not subject to errors present in the code and automatically generates code documentation in Confluence.End-to-end design of a frontend for creating model contracts, allowing scientists to insert model metadata. The frontend implementation was done using streamlit. The authentication of the contract captured in the frontend is done through the Databricks API, starting a job that executes and validates all fields.Implementation of the integration of Kraken data with OpenMetadata, through the development of a cloud application using Google Cloud Platform (GCP). I executed all phases of the project, from conception to final delivery, ensuring that Kraken data was integrated efficiently and reliably, using a scalable and resilient architecture, aligned with the bank's security and performance requirements.I took full responsibility for serving Kraken customers, answering questions and providing ongoing support regarding the product. I acted as a focal point between the MLOps team and clients, playing an active role in monitoring and executing client models, ensuring the delivery of personalized, high-quality solutions.

Responsible for implementing the computer vision system for the LYSA robot, a robotic dog for visually impaired personnel.I implemented a system within a Raspberry using ROS, where one of the nodes detects people and objects of interest through a convolutional neural network. When one of the target objects was identified in the robot's camera image, a message was published in a topic, where a second ROS node was subscribed. This node plays the role of checking and classifying the detected object according to its priority and the distance between the disabled person and the object. Based on this classification, it is possible to publish a new message on another topic, which allowed another node, when reading the published message, to trigger an audible warning, alerting the user about the objects around them.

I worked in two sectors of the company, being responsible for the entire IT sector of the factory.The main activities I carried out focused on IT, where I was able to develop a control system for equipment, hardware and software maintenance of systems connected to the network, auditing, factory server backup, improvement projects, among others. I also carried out activities such as acting as an engineering intern, carrying out planning for adjustments in the civil, electrical and hydraulic areas of the factory.

Development of the project ¨APP IFJF: A Virtual Guide with Augmented Reality¨, which aims to provide the user with a unique experience when visiting the physical facilities of our Campus. Using a smartphone, using its camera, Augmented Reality features are displayed. The company's operating system, iOs, and its programming IDE, XCode, were used. The language used for such programming is Objective-C.