Seminar on Large Language Models for Brazilian Portuguese

Registration is now open for the Seminar on Large Language Models for Brazilian Portuguese, which will take place on June 23–24, 2025, at the Roberto Salmeron Auditorium, Faculty of Technology (FT), Darcy Ribeiro Campus – Asa Norte, University of Brasília (UnB), in Brasília, Brazil – in-person event.

The seminar will bring together researchers, professionals, and representatives from the public and private sectors to discuss the advances, challenges, and opportunities related to the development and application of large-scale language models tailored to Brazilian Portuguese.

Main topics of the seminar:

• Development of LLMs adapted to the linguistic and cultural nuances of Portuguese;

• Collaboration between academia, industry, and government for ethical and inclusive AI;

• Practical applications in education, health, and public services.

Register here:
https://docs.google.com/forms/d/e/1FAIpQLSeoDSTXTgpR6n3G2QZUTRTAM7zT90RBX8xHryytMfUpZrs9uw/viewform

Program Overview

Monday, June 23

7:30 AM – Check-in

8:35 AM – Opening remarks – UnB, LNCC, NVIDIA, and MCTI

8:45 AM – “AI for the Good of All” – Brazilian Artificial Intelligence Plan (PBIA)
Henrique de Oliveira Miguel, Secretary for Digital Transformation Science and Technology

9:00 AM – NVIDIA’s Experience in Building Sovereign LLMs in Other Countries
Evan Acharya, Senior Solutions Architect – NVIDIA

9:30 AM – Roundtable: Universities & ICTs

• Laurent Dardenne – LNCC

• Anderson Soares – CEIA/UFG

• Altigran Silva – UFAM

• Fabio Cozman – CIAAM/USP

10:30 AM – Break & Poster Session

11:00 AM – Roundtable: Universities & ICTs

• Ricardo Ruviaro – CENIA/UnB

• Leandro Villas – H.IAAC/Unicamp

• Júnia Cristina Ortiz Matos – SENAI-CIMATEC

• Nicholas Corrêa – University of Bonn

12:00 PM – Lunch

2:00 PM – NVIDIA Platform for Building Sovereign LLMs
Pedro Mário Cruz e Silva, Senior Solutions Architect – NVIDIA

2:30 PM – Roundtable: Companies and Startups

• Cristiano Vargas – PIT/PI

• Juliana Crepalde – KUNUMI

• Sérgio Kamache – SERPRO

• Nelson Leoni – WideLabs

3:30 PM – Break & Poster Session

4:00 PM – Roundtable: Companies and Startups

• Bruno Pierobon – NeoSpace

• Marcellus Amadeus – Amadeus IA

• Luiz Henrique Bonifácio – NeuralMind

Poster Presentations:

• Carcará: Efficient Deployment of an Open-Source LLM with 685 Billion Parameters on the Brazilian Supercomputer Santos Dumont (LNCC)

• Scientific Extraction with olmOCR and LoRA: Results and Metrics (IIA-LNCC)

• Gypscie: Machine Learning Lifecycle Management with Artifact Tracking and Multi-Environment Execution (DEXL LAB – LNCC)

Poster application demonstrations will take place during the session.

Tuesday, June 24

8:00 AM – Hands-On Training
“NVIDIA Playbook for non-English LLMs”
Speaker: Pedro Mário Cruz e Silva, DSc
NVIDIA Senior Solutions Architect

9:45 AM – Coffee Break

10:00 AM – Hands-On Training (continued)
“NVIDIA Playbook for non-English LLMs”
Speaker: Pedro Mário Cruz e Silva, DSc
NVIDIA Senior Solutions Architect

12:00 PM – Lunch

The event is a partnership between the Artificial Intelligence Center of the University of Brasília (CEN.IA/UnB), the Artificial Intelligence Institute of the National Laboratory for Scientific Computing (IIA-LNCC), and NVIDIA, with support from the Ministry of Science, Technology and Innovation (MCTI).

Technical Committee
Ricardo Ruviaro (UnB)
Fabio Porto (LNCC)
Pedro Mário Silva (NVIDIA)

Organizing Committee
Anmily Paula Martins (IIA-LNCC)
Graziele Soares (IIA-LNCC)
Tathiana Tapajóz (IIA-LNCC)

Organizers
CEN.IA/UnB
IIA/LNCC
NVIDIA

Support
Ministry of Science, Technology and Innovation
IIA/LNCC
NVIDIA

IMPORTANT NOTICE
We inform you that all spots for the Seminar on Large Language Models for Brazilian Portuguese have been filled. A waiting list will be made available in case of cancellations or new openings. Interested individuals may sign up via the link below and will be notified by email:
https://docs.google.com/forms/d/e/1FAIpQLScp9GORzC8VbIVrjuD1Atug_RkWDJWhKZ4Bf8JKGkX_qMxY2g/viewform?usp=header
Waiting list criteria: first come, first served.