Tutorials

Tutorial WESAAC

Multi-Agent Systems based on Language Models

Date, time and room to be confirmed.

About the course

Recent advances in Large Language Models (LLMs) have transformed the way humans interact with computational systems. Unlike traditional chatbots based on predefined rules and rigid workflows, LLM-based agents can understand natural language, adapt to context, and generate coherent and flexible responses. These capabilities have enabled applications across domains such as education, healthcare, customer support, marketing, legal services, and human resources. More recently, LLM-based multi-agent systems have emerged as a promising paradigm for addressing complex problems. By coordinating specialized autonomous agents, these architectures can distribute responsibilities, improve scalability, and enhance solution quality. However, they also introduce challenges related to context management, interaction monitoring, and conversational workflow orchestration.

This introductory short course presents the fundamental concepts and practical aspects of building multi-agent systems with LLMs. The program is organized into three parts: (1) Introduction and Foundations, (2) Practical Technical Demonstrations, and (3) Hands-on Exercises and Discussion. No advanced background in Artificial Intelligence or Natural Language Processing is required; only basic Python programming knowledge is expected. Participants will receive complementary learning materials to support content review, practical experimentation, and adaptation of the examples to academic or professional projects.

The course will be delivered in Portuguese, although some frameworks, libraries, and technical documentation used during the demonstrations may be available only in English. The target audience includes undergraduate and graduate students in Computer Science, Computer Engineering, Information Systems, and related fields who are interested in developing applications based on LLM-powered agents. The course is limited to 30 participants, with registration through an online application process and participant selection by the organizers.

Instructors

Prof. Dr. Julio C. dos Reis

Institute of Computing, University of Campinas, Brazil

Dr. André Gomes Regino

CTI Renato Archer, Brazil

Inferência causal para tomada de decisão em sistemas complexos

Date, time and room to be confirmed.

About the course

A oficina apresenta os fundamentos da inferência causal e demonstra como técnicas modernas permitem estimar efeitos causais mesmo quando existem variáveis não observáveis, um dos principais desafios enfrentados em ambientes reais.

Instructor

Eduardo Laurentino

Itaú

Construindo sistemas inteligentes com visão computacional generativa

Date, time and room to be confirmed.

About the course

Mostrar como modelos modernos entendem documentos, imagens e vídeos. Os participantes experimentam pipelines de Computer Vision e modelos multimodais para resolver problemas de negócio.

Instructor

Miguel Wanderley

Itaú

Cidades Agênticas: Um Control Plane de Resposta a Crises com Agentes Autônomos na AWS

Date, time and room to be confirmed.

About the course

Enchentes, apagões e vazamentos de gás exigem decisões em minutos, mas as entidades de resposta operam em silos. Nesta sessão de 4 horas, você vai conhecer o Agentic-City: um control plane onde agentes autônomos detectam crises a partir de fontes públicas de dados, enriquecem o contexto com APIs governamentais e dados geoespaciais, debatem cenários de impacto (trânsito, defesa civil, hospitais, escolas) e recomendam ações a um operador humano, responsável pelas decisões e acompanhamento das ações de remediação e seus impactos. Após 1h de conteúdo e demonstração do Agentic-City, você colocará a mão na massa (3h) usando Amazon Bedrock e AgentCore para conhecer as principais tecnologias utilizadas para implementar agentes de IA de forma segura, escalável e governada.

Instructors

Cleber Gomes da Silva

Arquiteto de Soluções Senior da AWS para Governo; ex-aluno de Ciência da Computação da UFMT

Vinicius Soares Batista

Arquiteto de Soluções Senior da AWS para Governo