Science moves forward through people and places

I presented research on interpreting deep neural networks and answered questions from the audience at the Manchester summit.

The audience’s questions reminded me that a model’s result is more useful when people can make sense of how it arrived there. That is why interpretability belongs in the research conversation.

At the University of Manchester DigiLab, I introduced school-age visitors to machine learning through movement. Using Kinect motion-sensing cameras, participants could see how movement data lets a computer study walking patterns, then try the demonstration themselves.

I enjoyed making the research something visitors could explore through their own movement. Once they began using the technology, their questions showed how quickly an abstract idea could become tangible.

I represented the University of Manchester at its Industry 4.0 Summit stand and welcomed visitors to explore the university’s robotics and machine-learning work, including a team robotics demonstration.

Taking research out of the lab made the work tangible: people could see a system, ask what it could do, and connect technical ideas with industrial practice.

The Alan Turing Institute, the UK’s national institute for data science and artificial intelligence, organized the Fairness in AI working group I joined in London during a week of data-science work. Researchers gathered around a shared table to work through questions about fairness in AI.

AI systems can shape people’s opportunities and access to services, so fairness cannot be treated as a final check. It requires people with different expertise and experience to ask who could be overlooked and how a system’s effects can be evaluated. Working through those questions together is part of responsible engineering.

I spoke about artificial intelligence and neuroscience in a Cambridge Science Festival program at the Massachusetts Institute of Technology (MIT). The audience included people interested in how computational methods can help researchers ask new questions about the brain.

The discussion made interdisciplinary research feel concrete: different ways of studying the brain can inform one another when people share their questions in the same room.

I helped organize the AI Latin American SumMIT at MIT Media Lab to make room for researchers and policymakers from the region to shape conversations about artificial intelligence around Latin America’s own needs and experience. The January 21–23 summit brought people together for panels and exchange; a separate two-day hackathon followed on January 24–25. Speakers Pablo Samuel Castro and Andreas Müller led a machine-learning workshop, and I led hands-on sessions on deep learning, neural networks, and PyTorch, a software library for building machine-learning models. We worked through how models learn from data and supported teams as they turned their ideas into projects.

I hoped the summit would help create connections that lasted beyond the week: researchers could share practical methods, teams could build together, and participants from Latin America could steer the discussion toward regional priorities. That kind of exchange gives more people a part in deciding what AI should be used for and whom it should serve.

At Massachusetts General Hospital’s Future of Mental Health Conference, I met Jim Yong Kim and heard him speak about global mental health and access to care. A physician and anthropologist, Kim co-founded Partners In Health, led the World Health Organization’s 3 by 5 initiative to expand HIV treatment, and served as president of the World Bank Group from 2012 to 2019. His career has linked health, communities, and the institutions that shape who can receive care.

Mental health research matters because it shapes the lives of people seeking care, their families, and their communities. Kim’s talk left me with a question I carry into my own work on AI and health: can new knowledge reach the people who need it, and can the systems around them make that possible?

We hosted Demis Hassabis at MIT on April 5, 2022, for “Using AI to Accelerate Scientific Discovery,” and I was glad to be part of the hosting team. A central question was how AI might give researchers new ways into scientific problems that have resisted familiar approaches. Hassabis later shared the 2024 Nobel Prize in Chemistry for computational methods that predict the three-dimensional structures of proteins, the molecules that carry out much of life’s work.

That connection between computation and biology inspires me: when a method is built around a question that matters, it can help scientists see a problem differently and choose what to investigate next. The people and ideas in the seminar made the distance between a computer model and a living system feel smaller.

I co-organized the MIT workshop and co-led a session on neurosymbolic programming for science: combining the pattern-finding strengths of machine learning with the explicit structure of programs and scientific knowledge.

This workshop gave me a chance to help shape the question and work alongside people exploring how programming can support science.

At the Harvard University Mexico Conference, I joined conversations about how technology and financial innovation could respond to Mexico’s opportunities and needs.

The conversation brought innovation back to place and purpose. New technology has to meet the economic and social realities of the communities it is meant to serve.

I attended Finale Doshi-Velez’s Broad Institute talk on interpretable machine learning. The talk considered when models can be understood well enough to support oversight and insight, a question that remains central to using AI responsibly in science and health.

The session returned me to a question I carry into my own work: an AI result is more useful when people can examine what shaped it and decide whether it deserves their trust.

I attended the conference and discussed my group’s research with fellow researchers in the poster session.

The exchange placed our work alongside a wider effort to use AI in cancer research and care. It reminded me that better computational tools matter when they help people understand disease and develop better treatments.

I helped host Stéphane Bancel, Moderna’s chief executive at the time, for a conversation with Andrew Lo, an MIT Sloan professor of finance. Lo directs the Laboratory for Financial Engineering, where he applies quantitative methods to questions in finance and biotechnology. Together we considered how advances in biotechnology might shape medicine, and how financing can help research move toward patients.

Having a biotech leader and a financial economist in conversation made the path from discovery to treatment visible as a shared problem. Scientific insight, clinical need, and the resources to develop a therapy all have to meet before an idea can reach patients.

In April 2025, when Aravind Srinivas visited MIT, I joined the audience at the Martin Trust Center for MIT Entrepreneurship for his conversation with Bill Aulet. As a founder who uses the center’s spaces, I was glad to welcome him to MIT.

I hoped he would find MIT’s entrepreneurship spirit inspiring. It is the same community of fellow founders I work alongside at the Martin Trust Center.

Right after Yann LeCun’s keynote at MIT Technology Review’s EmTech conference at the MIT Media Lab in November 2025, I reconnected with him. We talked about where the next wave of AI is heading: beyond deep learning, toward neurosymbolic intelligence that can reason, explain, and truly understand.

Every conversation with Yann reminds me how curiosity drives progress, and how much of the AI frontier is still unwritten.

At Harvard’s 2025 Thich Nhat Hanh Center mindfulness symposium, I shared a table with Jon Kabat-Zinn and asked what excited him about mindfulness and what he thought might come next. A molecular biologist, Kabat-Zinn founded the Stress Reduction Clinic at the University of Massachusetts Medical School in 1979 and developed its Mindfulness-Based Stress Reduction (MBSR) program.

Our exchange connected Jon Kabat-Zinn’s decades of scientific and clinical work on mindfulness with a practical question about everyday attention and well-being. I left thinking about how a field evolves: through careful study, practical application, and continued openness to what experience can teach.

I joined the interdisciplinary team that organized the 2022 Neurosymbolic Programming Summer School at Caltech. Students explored how machine learning can find patterns while symbolic programs make relationships and constraints explicit, a combination that may help computers support scientific reasoning.

I helped create the school because these questions need more than a single discipline. When students, engineers, and scientists can test ideas together, they can ask not only what a model predicts but whether its reasoning is useful and understandable to the people doing science.

At Stanford’s Drug Discovery Symposium, I met chemist Carolyn Bertozzi and we spoke about health impact and research opportunities. Her development of click chemistry and bioorthogonal chemistry, which lets researchers study biomolecules in living systems without disrupting their normal chemistry, earned her the 2022 Nobel Prize in Chemistry.

I was inspired by how a new chemical method can open a line of inquiry that basic research could not reach before. Our exchange returned me to a question behind translational science: how do we carry discovery toward better ways to understand and treat disease?

I co-organized minisymposium 261, a focused research session titled “Artificial Intelligence and Neuroscience: From Neural Dynamics to Artificial Agents,” and presented talk 261.02, “Understanding neural dynamics under flexible visuomotor tasks with neural decoding and manifold learning.” Neural decoding uses brain recordings to infer behavior; manifold learning helps reveal patterns in complex data. The Society for Neuroscience program listed the talk for October 21 at 8:35 a.m. in Room S406A.

Co-organizing the session let me help build a place where neuroscience and AI could challenge and inform one another.

During New York TechWeek in June 2024, I introduced Equ Healthcare and presented its health-technology product to the TechWeek community. The launch brought the venture and its product into conversations around technology and healthcare.

The presentation made the next question tangible: how can a health-technology idea respond to the cost of care and help make healthcare more accessible?

I joined the AI for All panel at the 2025 Congressional Hispanic Caucus Institute Tech Summit to discuss building an inclusive future in health technology.

For health AI to serve people well, trust, access, and accountability have to be part of the conversation from the beginning. This panel brought those questions into the room with people working across research, technology, and policy.

At Innovation Match MX in Mexico City, I presented research modeling gait, walking patterns, over time and space from floor-sensor data, using deep learning and other new machine-learning methods. The forum brought research into a wider conversation about technology, institutions, and innovation in Mexico.

Sharing the work in Mexico connected my research in sensing and machine learning with people thinking about how new ideas can contribute to healthcare and the country’s innovation ecosystem.

At Tecnológico de Monterrey’s Santa Fe campus, I led a masterclass on artificial intelligence and programming techniques for science.

The students’ questions turned the session toward problems from their own fields. Working through those examples showed how computational methods become useful when they help answer a question that matters to the person asking it.

I spoke with the first student cohort at Aporta Perú’s Institute for Advanced Analytics and Data Science in Lima. The gathering brought students and educators together around the possibilities of data science and machine learning.

I left encouraged by the cohort’s shared ambition to use data science in practical, locally meaningful ways. Teaching is also a conversation: students’ questions show where a field can go next.

I joined international peers to share recommendations on AI strategy at the World Artificial Intelligence Conference Governance Forum.

The forum widened the question from what AI can do to how institutions can guide its use, weigh its effects, and make responsible choices.

I presented footstep sensing research at the IEEE SENSORS Conference in Busan and received the Best Student Paper Award in Optical Sensing Applications. The conference brought researchers together to share advances in sensors and their use.

The award encouraged me to keep asking how signals from the world around us can become useful knowledge. It also gave me a chance to share the work with researchers developing new sensor technologies.