As ML Platform Engineer @ Feature Platform, my role includes developing and implementing new strategies for efficient data processing, storage, and low-latency data availability at a terabyte scale in both batch and real-time scenarios. Here are some key highlights:• Developed a Go Lang-based REST API, a core component of the Feature Store, enabling low-latency feature retrieval up to 30K RPS with 12ms P99 latency.• Refactored legacy Python SDKs used by data science teams to build and test their ML features.• Refactored a custom RedisSink for Apache Spark to handle upsert operations at scale (~500K writes/s).• Developed a system for monitoring and notification of feature freshness for batch and stream features.• Developed an event-driven artifacts metadata management service to automate workflows (source tables, jobs, and features) and decoupling architectural components.Additionally, I serve as a technical reference for the team, providing expertise in tools such as Python, Go Lang, design patterns, and software architecture. I have extensive experience with a wide range of tools, but the ones I use daily include: Python, SQL, Go Lang, Scala, Apache Spark, Apache Kafka, Apache Airflow, Delta Lake, PostgreSQL, Redis, AWS S3, Docker, Kubernetes, Prometheus, Grafana, Datadog, Databricks, AWSAs ML Platform Engineer @ GenAI Platform (04/2023 ~ 09/2023):• Designed and built a Generative AI Platform to provide controlled, scalable, and fault-tolerant access to foundation model providers like OpenAI, AWS and open-source alternatives.• Designed and built a load balancing service based on custom heuristics to efficiently forward requests to the appropriate GPT models, consequently reducing costs and increasing tolerance to errors (e.g., rate limits, server overload).

As Machine Learning Platform Engineer @ Feature Platform:• Refactored legacy components used for processing data to improve maintainability, availability, performance, and cost-efficiency.As a Machine Learning Engineer @ Model Management Project:• Designed and built a solution to automatically collect and present ML model KPI results to stakeholders.• Designed and built scalable and cost-efficient data pipelines to collect ML model metadata and track the complete ML model lifecycle.• Improved ML model tracking solutions to be more fault-tolerant and self-recoverable.

• Search and analysis of MLOps tools, in order to identify the maturity and usability of the tool for client projects.• Design and development of robust and scalable automated machine learning pipelines (including model deployment & serving steps).• Implement MLOps practices in client projects for developing and improving scalable ML-powered applications.• Development, improvement and evaluation of machine learning models in the telecommunication field.

• Design, development and maintenance of machine learning pipelines (including model deployment & model serving) so that they are robust, scalable and easy to operate.• Development, improvement and evaluation of machine learning models (e.g., customers segmentation, points of sale segmentation, demand forecast, stock prediction, etc.).• Development of an internal platform that encapsulates the products and services offered by the company, enabling data scientists to be able to meet demands quickly and objectively.• Exploratory data analysis for customers to extract information and generate insights.

• Research and develop projects related to the classification of non-coding RNAs using deep learning strategies in all stages of the development cycle, from data wrangling to the implementation and evaluation of the built algorithms• Conducted a study on the impact of using different hyperparameters on the performance of deep feedforward networks for the classification of PIWI-Interacting RNAs (piRNAs). The study resulted in the paper "Performance Analysis of Deep Neural Networks in piRNAs Classification" published at the Encontro Nacional de Inteligência Artificial e Computacional (ENIAC) 2018.• Developed a state-of-the-art classifier whose accuracy is up to 9% greater than the best predictors in the literature, such as piRNN and piRNAPredictor, with a computational cost approximately 8 times lower. These results are being reported by articles published in scientific magazines, newspapers and conferences.

• Defined the schedule, activities and themes present at the event.• Performed the search and invitation of contacts and professionals responsible for providing the activities defined for the event.• Organized the Desafio de Programadores (Programmers Challenge Contest) at the 10th UFSCar Computing Week.

Biologists work daily with a huge volume of data. Keeping data organized, analyzing it and extracting relevant information is complex work. Thus, the computer programming becomes an important tool to handle such tasks, enabling a robust and elegant resolution of various problems of analysis and data organization.In order to disseminate this knowledge, I participated in the organization of the Python Workshop for Biological Data where lectures are offered, classes of introduction to programming with the Python language and several practical applications of analysis and processing of biological data, of which I was lecturer.