Pedro Spinosa

ML Engineer @ Nubank | AI Platform | MLOps & Infra

Brazil

Get in Touch View Experience
Pedro Spinosa

About Me

Software Engineer with 3 years of experience in Machine Learning infrastructure and platform engineering, specializing in building SDKs, workflow orchestration tools, and scalable model deployment services. One of the main contributors to Nubank’s internal ML workflow SDK, enabling all DS/ML teams to run training and scoring pipelines on Kubeflow. Experienced in managing large‑scale Kubernetes clusters, serving both real‑time and batch models, and covering the entire ML lifecycle from experimentation to production. Passionate about simplifying ML operations, collaborating with ML engineers and researchers, and creating reliable, testable, and maintainable platform solutions.

4

Years Experience

21

Skills

2

Certifications

Experience

Machine Learning Engineer

Nubank

September 2024 - Present (11 months) São Paulo, São Paulo, Brazil
  • Built an integration service that enables third-party model deployment within the current infrastructure, supporting the addition of state-of-the-art LLM models and automating the onboarding workflow for external models
  • Provided ongoing support and enhancements to the current ML development tool, adapting it to meet evolving use cases and optimizing the model training and scoring workflows
  • Developed technical reference materials and documentation for AI infrastructure, conducted tool training for development teams, and maintained ongoing feedback loops with users to continuously improve MLOps solutions
  • One of the main contributors to the internal SDK for authoring and orchestrating Data/ML workflows, enabling all DS/ML teams to design, run, and scale workflows seamlessly on Dagster and Kubeflow

Junior Machine Learning Engineer

Nubank

January 2024 - August 2024 São Paulo, São Paulo, Brazil
  • Implemented a lifecycle process for Python versions, enabling reliability of models and libraries, along with updating all AI platform libraries to comply with updated Python range versions
  • Led performance optimization initiatives, achieving significant latency reductions in model inference, enabling multi-parallelism and scoring in batch
  • Enhancing our model development tool to better align with user experience requirements, while also strengthening foundational processes to improve the overall development experience—particularly for new users entering the system

Machine Learning Engineering Intern

Nubank

September 2022 - December 2023 São Paulo, São Paulo, Brazil
  • Developed standardized containerization patterns for ML model deployment, improving resource utilization
  • Built foundational REST API endpoints for model serving, streamlining deployment workflows
  • Contributed to team knowledge sharing, enhancing documentation and best practices in ML infrastructure

Machine Learning Researcher

Insight Data Science Lab

May 2021 - August 2022 Fortaleza, Ceará, Brazil
  • Developed an application that utilizes Named Entity Recognition to identify and highlight entities within a text‑based bulletin. To achieve this, the API utilizes two libraries/frameworks for building base models: SpaCy and Keras
  • Developed an ML operation tool implemented as a class enabling parallel or queued machine learning model training, resulting in a productivity increase of at least 50% by utilizing background threads
  • Created a testable and reliable monitoring class using WebSockets, providing real‑time updates on machine learning model training progress and performance during each epoch for improved ML operations visibility

Skills & Expertise

AI/ML

Dagster Deep Learning Kubeflow Machine Learning NLP/NER

DevOps

CI/CD Containerization Docker Kubernetes MLOps

Cloud

AWS

Programming

C/C++ Clojure Go(learning) Python Rust(learning)

Soft Skills

Documentation Team Collaboration

Backend

FastAPI REST APIs

Engineering

Performance Optimization

Projects

AXL Workflows

AXL Workflows

OSS WIP
PyPI version Python versions

AXL Workflows (axl) is a lightweight framework for building data and ML workflows with a class-based Python syntax. Build a workflow once, then run it locally or on Argo/Kubeflow/Dagster/Airflow.

Python Rust Dagster Argo Workflows Kubeflow Airflow Kubernetes ML Data Engineering

Education

Bachelor's degree, Computer Science

Federal University of Ceara

2019 - 2023

Fortaleza, Ceará, Brazil

Certifications

AWS Academy Machine Learning Foundations

AWS

AWS Academy Cloud Foundations

AWS

Get in Touch