Fernanda Viegas
Gordon McKay Professor of Computer Science, Harvard SEAS; Sally Starling Seaver Professor, Harvard Radcliffe Institute for Advanced Study; Chandra Family HBS University Fellow
Artificial Intelligence in Brazil Panel Series- Day 1
Date and Time
April 17, 2025
09:30AM - 05:15PM EDT
Location
CGIS South, S020 Belfer Case Study Room
Join us for this two-day panel series focusing on cutting-edge AI research led by Brazilian experts while exploring broader discussions on regulation, policy, law, digital rights, privacy, workforce implications, and education in the Brazilian context. Speakers will represent diverse backgrounds and institutions, including academia, government, and the private sector.
Panel Series organized by Flavio Calmon, Caio Vieira Machado, Claudio Verdun, and Lucas Monteiro Paes.
Presented in collaboration with the Center for International Development, Center for Research on Computation and Society, Weatherhead Center for International Affairs, Harvard Data Science Initiative, and the Consulate General of Brazil in Boston.
Featured Speakers
Cicero Nogueira dos Santos
Research Scientist, Google DeepMind
Sandra Avila
Assistant Professor, Institute of Computing, UNICAMP
Rafaela Nicolazzi
Data, Privacy & Consumer Protection Lead, Global Affairs - OpenAI
Eugenio V. Garcia
Director, Department of Science, Technology, Innovation and Intellectual Property, Ministry of Foreign Affairs of Brazil
Mariana Valente
Assistant Professor at the University of St. Gallen Law School, Director of InternetLab
Flavio Calmon, Associate Professor, Harvard SEAS will deliver the opening remarks.
Flavio Calmon is an Associate Professor of Electrical Engineering at Harvard's John A. Paulson School of Engineering and Applied Sciences. Before joining Harvard, he was a social good post-doctoral fellow at IBM Research in Yorktown Heights, New York. He received his Ph.D. in Electrical Engineering and Computer Science from MIT. His main research interests include information theory, signal processing, and machine learning.
Flavio Calmon's research group focuses on information theory, signal processing, and machine learning, with the goal of developing foundational theory to guide the design of trustworthy machine learning algorithms. Recently, their work has concentrated on three critical challenges in machine learning: fairness, privacy, and reliability. Their vision is that information theory is essential for the responsible design of machine learning systems and that a theory-guided approach can significantly outperform heuristic methods. More details about their research can be found in the selected publications below.
Calmon considers himself a scientist with an engineering mindset, enjoying fundamental research that drives practical applications. His approach follows an information-theoretic blueprint: rather than asking, "How can a given system be improved?", he inquires, "What is the absolute best a system can achieve?" By using information theory to characterize the fundamental performance limits of machine learning systems, he designs algorithms aimed at achieving these limits.
As a Brazilian-American engineer, Calmon is passionate about broadening the participation of students from diverse backgrounds, including those from Latin America, in information theory and machine learning. He is also enthusiastic about addressing data-driven challenges specific to the developing world. Those interested in these areas are encouraged to reach out.
His research group is supported by the NSF, as well as generous gifts and awards from Google, IBM, Amazon, and Oracle Research.
Skin cancer is a major public health problem that could benefit from computer-aided diagnosis to reduce the burden of this common disease. While deep learning algorithms achieve "superhuman" performances in different contexts, models struggle to generalize, stalling deep learning adoption in critical tasks like skin cancer analysis. Our research group has been working on skin image analysis (classification, synthesis, and debiasing) for over a decade. In this talk, I will delve into our approaches and achievements — until I discovered I was collaborating on a Eugenics skin cancer project and immediately redirected our focus. I will discuss our ongoing efforts at developing a solution capable of effectively generalizing across a diverse population with darker skin tones.
Sandra Avila is an Assistant Professor at the Institute of Computing, University of Campinas (UNICAMP). She holds a double degree Ph.D. in Computer Science from the Federal University of Minas Gerais (UFMG) and Sorbonne Université (UPMC Sorbonne, Paris 6, France) in 2013. Her research focuses on Artificial Intelligence for Social Good, with an emphasis on healthcare and sensitive media analysis. She has received several prestigious awards, including the Google Latin America Research Awards (2018–2021), Google Awards for Inclusion Research (2022), and the UNICAMP-Instituto Vladimir Herzog Academic Recognition Award in Human Rights (2022, 2023). In 2020, she was selected by the Brazilian Academy of Sciences to represent Brazil at the BRICS Young Scientists Forum in the field of Artificial Intelligence. In 2022 and 2023, she was recognized among the top 2% of the most influential scientists worldwide, according to Stanford/PlosOne/Elsevier rankings.
Claudio Mayrink Verdun is a mathematician working with the mathematics of AI and machine learning at Harvard’s School of Engineering and Applied Sciences under the mentorship of Flavio Calmon. His research focuses on trustworthy machine learning, exploring concepts such as fairness and arbitrariness, as well as mechanistic interpretability techniques for large generative models. In particular, he concentrates on how parsimonious models, particularly sparsity, can be leveraged to enhance the understanding of machine learning. Over the years, Verdun has utilized optimization, statistics, and signal processing techniques to advance both the theory and practice of artificial intelligence. His work includes developing provably fast and scalable methods for machine learning models, creating uncertainty quantification techniques for high-dimensional problems involving large datasets, and understanding the discretion of large language models (LLMs). He is also passionate about applying these advanced techniques to practical domains, such as medical imaging (particularly MRI) and the use of AI for education.
Verdun had the privilege of completing his Ph.D. in mathematics under the guidance of Felix Krahmer within the Optimization and Data Analysis group, while concurrently being affiliated with the Information Theory group under the leadership of Holger Boche at the Technical University of Munich.
In this session, early-career researchers will present some of their innovative AI research.
Born in São Paulo, Brazil, Yan Kaled is the Founder and CEO of HiSolver, and holds a MEng in Computational Science and Engineering by Harvard SEAS, where he was awarded the the Leadership Fellowship from Estudar Foundation, and the Lemann and Behring Fellowships. At Harvard, Yan worked on the Master's Thesis that originated HiSolver, an AI-powered Social Marketplace that allows creators to scale their community building and digital businesses with a new suite of digital products, such as multimodal interactive AI Clones, and automatically generated Micro SaaS.
Matheus Farias is a 26-year-old Electrical Engineering Ph.D. student at Harvard University, advised by Prof. H. T. Kung. He is from Recife, Brazil, and has a keen interest in developing innovative solutions for real-life problems using interdisciplinary engineering approaches, particularly through the combination of machine learning, the Internet of Things, and hardware design.
Matheus graduated as valedictorian with a BSc in Electronics Engineering from the Federal University of Pernambuco (UFPE). During his time at UFPE, he worked in four different labs under five distinct faculty members. This diverse experience provided him with a unique perspective on multidisciplinary research and collaborative efforts, allowing him to effectively contribute to various research lines and environments. He is fortunate to have received five international awards, along with several national accolades, including the title of MIT Innovator Under 35 for outstanding science and technology innovators under the age of 35.
In addition to his academic work, Matheus is a member of the rock band Matucana, where he plays the keyboards. The band's first album was released in 2023. He also nearly became a professional badminton player, earning over 10 medals in state tournaments and University Games.
Claudio Mayrink Verdun is a mathematician working with the mathematics of AI and machine learning at Harvard’s School of Engineering and Applied Sciences under the mentorship of Flavio Calmon. His research focuses on trustworthy machine learning, exploring concepts such as fairness and arbitrariness, as well as mechanistic interpretability techniques for large generative models. In particular, he concentrates on how parsimonious models, particularly sparsity, can be leveraged to enhance the understanding of machine learning. Over the years, Verdun has utilized optimization, statistics, and signal processing techniques to advance both the theory and practice of artificial intelligence. His work includes developing provably fast and scalable methods for machine learning models, creating uncertainty quantification techniques for high-dimensional problems involving large datasets, and understanding the discretion of large language models (LLMs). He is also passionate about applying these advanced techniques to practical domains, such as medical imaging (particularly MRI) and the use of AI for education.
Verdun had the privilege of completing his Ph.D. in mathematics under the guidance of Felix Krahmer within the Optimization and Data Analysis group, while concurrently being affiliated with the Information Theory group under the leadership of Holger Boche at the Technical University of Munich.
Lucas Monteiro Paes is an AI researcher and mathematician. He is an Applied Mathematics Ph.D. candidate at Harvard University, working with Prof. Flavio Calmon. Previously, he was a Student Researcher at Google DeepMind in the Gemini Safety Team and an AI Research Scientist Intern at IBM Research at the IBM T.J. Watson Research Center.
Monteiro Paes uses theoretical insights to develop safe and trustworthy AI and machine learning systems. His research is driven by the belief that AI and ML systems should not only be accurate and efficient but also transparent, fair, and aligned with human values and societal norms. His research is supported by the 2024 Apple Scholars in AI/ML Fellowship.
Before joining Harvard, Monteiro Paes earned an M.S. in Computational Mathematics and Modeling from Instituto de Matemática Pura e Aplicada (IMPA), a renowned mathematics institute in the Tijuca National Park in Rio de Janeiro, Brazil.
Marcia Castro is Andelot Professor of Demography, chair of the Department of Global Health and Population, director of the Brazil Studies Program of the David Rockefeller Center for Latin American Studies (DRCLAS) at Harvard University, associate faculty of the Harvard University Center for the Environment, and faculty member of the Harvard Center for Population and Development Studies.
Her research focuses on the development and use of multidisciplinary approaches to identify the determinants of infectious disease transmission in different ecological settings to inform control policies. She has more than 20 years of research experience in the Brazilian Amazon, with a strong record in conducting household surveys and thorough knowledge of the local culture. Furthermore, she has more than 15 years of collaboration with Brazilian researchers, health secretariats, and the ministry of health, particularly related to infectious diseases.
Currently, Castro has projects on malaria, COVID-19, arboviruses, infant/child mortality and development, and climate change in the Brazilian Amazon. Specifically, she has been assessing the spatiotemporal pattern of COVID-19 spread in Brazil, mortality and fertility changes due to the pandemic, risk factors for mortality, and vaccine effectiveness.
When you interact with a chatbot, what does it “think” about you? Does it care about your age, level of education, or socioeconomic status? It turns out the answer is yes. Recent work in AI interpretability is beginning to provide intriguing answers about how implicit social cognition helps shape chatbot behavior. In this talk I'll describe an interface that translates the complex geometry of AI language models into a simple, understandable dashboard. This dashboard allows end users to see—and also control—how a chatbot perceives them. I argue that this type of transparency is an important step in helping people work with AI more effectively, safely, and enjoyably.
Fernanda Viégas is Sally Starling Seaver Professor at Harvard Radcliffe Institute and a Gordon McKay Professor of Computer Science at the Harvard John A. Paulson School of Engineering and Applied Sciences. She is also affiliated with Harvard Business School.
Viégas is a pioneer in data visualization and analytics known for her contributions to social and collaborative visualization through the creation of open-source tools. With her longtime collaborator, Martin Wattenberg, she coleads Google’s People + AI Research (PAIR) initiative, which advances the research and design of people-centric AI systems. Together, they founded the visualization studio Flowing Media, Inc., and led IBM’s Visual Communication Lab, where they created the public visualization platform Many Eyes.
Viégas earned her PhD in media arts and sciences from the Massachusetts Institute of Technology’s Media Lab. Her visualization-based artwork, created with Wattenberg, has been exhibited worldwide and is part of the permanent collection of the Museum of Modern Art, in New York.
Flavio Calmon is an Associate Professor of Electrical Engineering at Harvard's John A. Paulson School of Engineering and Applied Sciences. Before joining Harvard, he was a social good post-doctoral fellow at IBM Research in Yorktown Heights, New York. He received his Ph.D. in Electrical Engineering and Computer Science from MIT. His main research interests include information theory, signal processing, and machine learning.
Flavio Calmon's research group focuses on information theory, signal processing, and machine learning, with the goal of developing foundational theory to guide the design of trustworthy machine learning algorithms. Recently, their work has concentrated on three critical challenges in machine learning: fairness, privacy, and reliability. Their vision is that information theory is essential for the responsible design of machine learning systems and that a theory-guided approach can significantly outperform heuristic methods. More details about their research can be found in the selected publications below.
Calmon considers himself a scientist with an engineering mindset, enjoying fundamental research that drives practical applications. His approach follows an information-theoretic blueprint: rather than asking, "How can a given system be improved?", he inquires, "What is the absolute best a system can achieve?" By using information theory to characterize the fundamental performance limits of machine learning systems, he designs algorithms aimed at achieving these limits.
As a Brazilian-American engineer, Calmon is passionate about broadening the participation of students from diverse backgrounds, including those from Latin America, in information theory and machine learning. He is also enthusiastic about addressing data-driven challenges specific to the developing world. Those interested in these areas are encouraged to reach out.
His research group is supported by the NSF, as well as generous gifts and awards from Google, IBM, Amazon, and Oracle Research.
The pursuit of computer programs capable of understanding and generating human language has been a central driving force in the field of Natural Language Processing (NLP). Today, large language models (LLMs) demonstrate remarkable fluency and coherence in both comprehension and generation, a feat largely attributed to a relatively simple recipe: scaling up deep neural networks (DNN), massive datasets, and computational power. However, the path to this success was not straightforward. For years, the dominant paradigm focused on handcrafted rules, intricate feature engineering and (later) problem specific DNN architectures, with scale playing a less obvious role. This talk presents a personal journey through two decades of NLP evolution, tracing the shift from the painstaking art of representing words with hand-crafted features to the surprising, emergent abilities of today's LLMs.
Cicero Nogueira dos Santos is a Research Scientist at Google DeepMind, where he focuses on the research and development of large scale deep learning-based approaches for multi-modal data understanding and generation. He was a core contributor to the team that created Google's flagship AI models Gemini 1.5 and Gemini 2. Before joining Google, he worked as a Research Scientist at Amazon Web Services (AWS) and IBM Research. Over the course of his career, Cicero has proposed innovative machine learning approaches for natural language processing, computer vision and drug discovery. He holds a Ph.D. in Computer Science from the Pontifical Catholic University of Rio de Janeiro, Brazil.
Lucas Monteiro Paes is an AI researcher and mathematician. He is an Applied Mathematics Ph.D. candidate at Harvard University, working with Prof. Flavio Calmon. Previously, he was a Student Researcher at Google DeepMind in the Gemini Safety Team and an AI Research Scientist Intern at IBM Research at the IBM T.J. Watson Research Center.
Monteiro Paes uses theoretical insights to develop safe and trustworthy AI and machine learning systems. His research is driven by the belief that AI and ML systems should not only be accurate and efficient but also transparent, fair, and aligned with human values and societal norms. His research is supported by the 2024 Apple Scholars in AI/ML Fellowship.
Before joining Harvard, Monteiro Paes earned an M.S. in Computational Mathematics and Modeling from Instituto de Matemática Pura e Aplicada (IMPA), a renowned mathematics institute in the Tijuca National Park in Rio de Janeiro, Brazil.
AI is fueled by our personal data - can we control it? What protections do we have and should we have? This panel will explore the regulatory and ethical frameworks shaping AI governance and the Brazilian perspective on the tensions between privacy and AI development. This panel addresses various topics, including data protection, algorithmic accountability, and the balance between innovation and fundamental rights.
Eugenio V. Garcia is the former Deputy Consul General in San Francisco, Head of science, technology and innovation, and focal point for Silicon Valley. 30 years of professional experience in foreign policy and diplomacy. PhD International Relations. Academic researcher on artificial intelligence and global governance. Former senior adviser to the President of the United Nations General Assembly in New York, 2018-2020.
Rafaela Nicolazzi leads data, privacy, and consumer protection efforts for OpenAI's Global Affairs team in Europe, collaborating closely with product and legal teams and engaging with privacy and policy stakeholders and regulators. Before joining OpenAI, Rafaela spent 12 years at Google, where she led AI data, privacy & safety initiatives for the Government Affairs & Public Policy team.
Mariana Valente is an assistant professor of international economic law at the University of St.Gallen and a founder and director of InternetLab in Brazil. A law researcher at the intersection of law, technologies, and social sciences, she works on AI and platform governance, copyright and access to knowledge in the digital environment, and online gender-based violence. She was one of the members of the Brazilian Senate’s Legal Experts Commission for Artificial Intelligence for drafting an AI Bill.
Caio Vieira Machado is a lawyer and social scientist focused on the intersection of law, technology, and policy. His interdisciplinary research addresses critical issues such as AI fairness, platform regulation, content moderation, and scientific disinformation. Currently, Caio is a fellow at Harvard's School of Engineering and Applied Sciences, collaborating with mathematicians on machine learning and fairness.
As a co-founder of Instituto Vero, Caio worked with Brazil's leading social media creators to combat disinformation and influence tech policy. He has presented his work at numerous international forums, including the UN OEWG, NATO StratCom, the Brazilian Supreme Court, the Mexican National Electoral Institute, and the AI Athens Roundtable. Caio holds degrees from the University of São Paulo, the University of Oxford, and the Sorbonne. He is now a Ph.D. candidate at Oxford, focusing on disinformation during the Covid-19 pandemic.
Have questions? Email Tiago Genoveze, Program Director, Brazil Studies Program.
The views expressed at this conference are the views and opinions of the speakers and do not represent the official view of DRCLAS or the University.