Paulo Guilherme Pinheiro Pereira

June 2025 - Present

Developed a platform for hosting machine learning challenges, enabling large-scale benchmarking and collaborative research across CERN teams.
Designed and maintained GitLab CI/CD pipelines to automatically build and push container images to Harbor, enabling automated image lifecycle management and reproducible deployments via Helm charts.
Adapted and deployed the platform to run on Kubernetes, enabling scalable and containerized user submissions with GPU support, persistent volumes, and queue management.
Managed a GitOps workflow using ArgoCD for declarative application deployment and synchronization on Kubernetes, managing application state across environments.
Authored and upstreamed Helm charts for all core components of Codabench, enabling Kubernetes-native deployments with persistent volumes, ingress, and containerized services — merged into the official open-source repository.

September 2024 - April 2025

Implemented an MLflow environment for model comparison across metrics, supporting model selection and lifecycle management decisions.
Developed tools to validate and standardize large-scale transcription corpora for three language variants (Brazilian Portuguese, European Portuguese, and Mexican Spanish) using regex and text processing techniques, ensuring dataset consistency and reproducibility across model versions.

September 2022 - September 2024

Fine-tuned deep learning models for keyword spotting using Python and TensorFlow, achieving Samsung's accuracy target for the project.
Built Kubeflow pipelines for scalable and reproducible model training on heterogeneous computing infrastructure.
Conducted systematic hyperparameter optimization and model evaluation across training runs, tracking performance metrics to support model lifecycle management decisions.

February 2024 - July 2024

Created exercise lists and conducted weekly meetings to support students enrolled in the course, under the supervision of Professor Dr. Evandro Ruiz.

August 2023 - December 2023

Created exercise lists and conducted weekly meetings to assist students enrolled in the course.
Activities funded by the Undergraduate Teaching Incentive Program (PEEG).

June 2024 - May 2025

Developed features and improvements for SlicerTutorialMaker, an automated tool for generating reproducible 3D Slicer training tutorials, under the supervision of Professor Dr. Luiz Otavio Murta.
Created multiple tutorials using the SlicerTutorialMaker extension to support outreach and education in medical image computing.

January 2023 - September 2023

Implemented multi-objective approaches for optimizing expanded genetic codes using evolutionary algorithms, under the supervision of Professor Dr. Renato Tinós.
Paper accepted at BRACIS 2023.

August 2021 - July 2022

Applied multi-objective approaches to optimize economic portfolios, under the supervision of Professor Dr. Renato Tinós.
Project funded by the Unified Scholarship Program (PUB).

Technical Student @ CERN