Andressa

As part of my research, I evaluate distributed architectures on AWS by running identical application code on EC2, ECS, and Lambda, first under matched provisioned capacity and then under deliberately mismatched capacity, so platform effects can be isolated from workload effects. I benchmark heterogeneous workloads such as data-ingestion APIs, tabular processing, and image resizing to find the exact thresholds where performance and cost tip from one service to another.

Building on my previously published papers and current studies, I implement full-stack observability using metrics, logs, and traces, so reliability can be measured rather than assumed in a genuinely distributed environment.

Through this empirical analysis, I map out optimal resource allocations by demonstrating when to leverage ECS and EC2 for low-latency, CPU-intensive tasks, and when to utilize Lambda for efficient event-driven processing.

My research establishes clear infrastructure thresholds for heterogeneous workloads:

• Ingestion API yields optimal performance on Amazon ECS, benefiting from the containerized environment’s lower and more consistent latency
• CSV Parser demonstrates higher economic and operational efficiency when offloaded to AWS Lambda’s event-driven, serverless execution model
• Image Processor performs significantly better on ECS and EC2, matching this highly CPU-intensive workload with stable, sustained resource allocation

I worked on the full lifecycle of a Conversational AI system, from model development to deployment in a production-like academic environment.

We performed fine-tuning of the language model and focused heavily on improving inference performance and system robustness, combining fine-tuning, prompt optimization, intelligent caching for frequent queries, and refinements to the inference pipeline.

I also contributed to robustness testing under load, monitored model behavior in multi-user scenarios, and documented the complete path from fine-tuning and evaluation through to live serving. It gave me practical exposure to the real challenges of adapting, optimizing and operating AI models where latency, reliability and observability are critical.

To further enhance the system’s accuracy and mitigate hallucinations, we integrated a Retrieval-Augmented Generation (RAG) architecture. This setup allowed the model to dynamically query external knowledge bases, ensuring that the conversational agent provided grounded, context-aware, and factually precise responses in real time.

In recognition of my technical commitment, core contributions, and dedication to the lifecycle of this project, I was awarded official institutional funding and a formal certificate of recognition by the University of Brasília (UnB) for delivering a high-performance, production-ready AI solution.

We developed Ardunina, an innovative and patented educational robotics platform created through the Meninas.comp extension project at the University of Brasilia. It was designed to address the significant gender gap in technology, offering a welcoming, practical introduction to robotics, electronics and programming, tailored specifically for girls from public schools.

Building on the Erekopapa model, we reimagined the robot with an aesthetic that would appeal to girls, and created a companion educational book, Create Your Own Ardunina.

Our workshops use an immersive RPG narrative featuring iconic women in science, guiding participants through the full engineering cycle, from physical assembly to code, while building technical skill and self-confidence in a collaborative environment.

I presented the paper Ardunina: a robotics learning environment for young girls at PAEE/ALE 2026, the International Conference on Active Learning in Engineering Education, held at Taisho University in Tokyo.

PAEE/ALE is the joint meeting of two long-running international networks, Project Approaches in Engineering Education and Active Learning in Engineering Education, whose communities work on project-based, problem-based and student-centered engineering teaching. The paper will be indexed in Scopus.

Together with the girls from the Meninas.comp project, I co-created the educational book Create Your Own Smart Garden, a hands-on guide introducing public school girls to technology, sustainability and automation by teaching them to build their own intelligent vegetable garden using Arduino, sensors and simple programming concepts.

The book offers clear step-by-step instructions, detailed explanations of each electronic component, and progressive programming activities, letting participants monitor soil moisture, automate watering systems and visualize real-time data. A garden was a deliberate choice, since it gives the sensors something alive to look after and makes the link between a line of code and the world outside it immediate.

We also designed and built the smart garden itself, a set of potted plants wired to an Arduino with soil moisture sensors and an automated watering system. That working prototype is what we brought into the workshops, so the girls could see one running before assembling and programming their own. Having the real thing beside the book turned the instructions into something they could touch, test and take apart.

The initiative increased the girls’ technical confidence, sparked strong interest in technology careers, and showed how engineering can be a powerful tool for both innovation and environmental care.

As Partnerships and Community Leader for the AWS Cloud Club at the University of Brasilia, I organized technical talks, workshops and cloud-focused events connecting students with industry practice in cloud computing. I also secured the partnerships and sponsorships that funded the club’s activities.

My role involved coordinating with speakers, promoting events across the university, and creating opportunities for students to learn about AWS services, cloud architecture and real-world applications, strengthening the local cloud community and encouraging students to explore professional paths in the field.

I was actively engaged in an extension project at the University of Brasília focused on fostering gender diversity in Computing. The initiative aimed to increase the participation of girls and women in the field through three main areas like robotics, competitive programming, and communication.

Within this project, my work was centered on robotics. I contributed by designing and developing educational prototypes using Arduino, as well as creating and facilitating hands-on workshops that introduced young girls to robotics concepts in an accessible and engaging way. I also collaborated with teachers to support robotics education in public schools across the Federal District and surrounding areas. Through these activities, the project sought to reduce educational barriers and encourage more female students to explore computing and engineering.