I'm currently working on cross-team initiatives that involve cloud platform integrations at the CDN level and enhancement of search-engine/database systems.

My work here was a collaborative effort, focusing on enhancing systems' cost-efficiency under high workload pressure and improving systems maintenance with automation and good practices. I worked on many teams, from a team designed to improve internal services to teams managing engineering observability or traffic concerns such as cache at the edge and a team responsible for designing the company's infrastructure architecture future. The following points highlight some of my achievements. - I maintained ownership of a search engine platform critical for the company's ecosystem, responsible for maintaining high resilience, delivering improvements/fixes, and leading the strategic planning for this service.---- After research/experimental investigation, I changed CPU architecture and resource usage to reduce the search engine's cost by up to 36% (in addition to the latter 48%). Initiatives involving data retention, o11y, and indexing management also reduced costs by 10%. The total cost reduction in this system was up to 70%.---- I upgraded the search platform to its latest version, enabling the service to run with both EC2-based infrastructure or via AWS-EKS under Helm charts versioning (and contributing to open-source in the process) to evaluate the best option. The upgrade improved the service performance overall (cutting CPU and disk usage by half) and enabled security/resiliency features that are only available in the latest versions.- Contributed to other teams' initiatives, writing automation with GitOps to refine settings version control, running infrastructure migration to improve observability with Prometheus instead of Cortex, studying via load tests trade-offs of a worldwide distributed infrastructure following by proposing paths to power up VTEX expansion towards Europe, and enabling growth teams with SRE support to speed up data importing of clients.

As Site Reliability Engineer, I focused on delivering cost-efficient applications and improving the resilience of systems efficiently.- Assumed ownership of a search engine platform critical for the company's ecosystem, being responsible for recovering it during incidents, and delivering required improvements/fixes.- Managed to increase observability, monitoring, and resilience, reducing incident occurrences from weekly to close to zero by investigating root causes and proposing/executing solutions.- Reduce the cost of the open-source search platform in e-commerce usage by about 48% with index refactor and EC2 instance type adjustment whilst the service usage more than doubled.- Delivered reliable, scalable, and cost-efficient EKS infrastructure and Helm deployment of open-source real-time database designed for modern analytics applications.

In my last contribution to this project, I have participated as co-author in the writing process of an international paper publication that studied the impact of the concurrency of indexing batches in the Solr system. I have also contributed to improving the documentation of the workload distributed replay tool (designed to deal with massively dense loads) and created standard deployment scripts.

I had the opportunity to work aside of VTEX's engineers that are outside of the VTEX Lab scope in order to accomplish a goal of interest for both groups. In this task, I managed to start a technology migration, upgrading the used Solr from version 6.6.6 up to 8.6.2. During this work I developed tools to easier reindexation process and search quality evaluation. I also had to plan and execute several experiments using a heavy workload on robust clusters in order to understand the impact of this version transition. Lastly, I lead the onboarding of the new time that assumed the project afterwards.

In the Acelera VTEX project under a partnership between SPLab@UFCG, LSD@UFCG, and VTEX company, we executed scientific methods to tackle optimization problems, design solutions, and propose infrastructure changes, aiming to increase performance and reduce resource usage (and consequently, reduce costs). In this project, we faced several branches of performance investigation, including impact on Apache Solr from the garbage collector, Solr's segment settings, and segment merges procedures. We also developed a robust workload replay tool able to effectively replay a given workload. I have contributed to the development of tools for metrics extraction and workload reproduction, and to researches tracks related to GC settings, and Solr's commits and merge.

Evaluated the feasibility of GCI on the Serverless application context through simulation, studying its impact on the response time of requests, the resource usage cost, and efficiency. The simulation modeled a FaaS architecture and perform a benchmarking of it under a Poisson's ratio workload. The results showed a response time improvement of up to 10.86% at 99.9th percentile, a reduction of costs by 7.22% at user perspective, and also indicated at the side of the platform providers an improvement of resource utilization by 14.52%.

Run a PIBIC research focused on conjecturing the impact of garbage collector interventions in the context of Serverless applications, precisely the FaaS group ones. This study showed that requests impacted by collector interventions have their service time increased by up to 8 times on tail.

Assisted in research about GCI, a request interceptor that improves the service time of cloud services by controlling Garbage Collector interventions, and their impacts. The evaluation of the experiments attained shown an improvement in tail latency of up to 47% by GCI's adoption.