Collaborating around AI for Global Health

My journey to founding NAAMII began with dissatisfaction during my engineering studies in Nepal. Despite topping a competitive entrance exam and attending the country’s best engineering school, I found the education focused on passing exams, with little emphasis on problem-solving or leadership. I envisioned an institution that would empower Nepali youth with global perspectives and leadership skills. This vision grew stronger while leading our student robotics team to international competitions. Competing against advanced teams from Japan and China, I saw our talent but recognized the lack of infrastructure and mentorship needed to institutionalize knowledge and drive consistent improvement. This gap made me question why Nepal wasn’t contributing to the global science and technology landscape.

After completing my degree, I co-established an innovation lab at my engineering school with the help of a professor. However, the lab faced structural and cultural resistance from faculty, reinforcing my realization that I needed deeper research expertise to drive meaningful change. Determined to return with impactful skills, I pursued advanced studies in France, promising myself I would be back in Nepal within 10 years.

In France, I met other Nepali students who shared my aspirations for advancing education and research in Nepal. Our experiences abroad exposed us to Europe’s research ecosystem, including institutions like INRIA and Max Planck. Their models of autonomous research groups and flexible partnerships with universities inspired us. Together with a surgeon in Nepal, we envisioned an institute in Nepal that could operate independently of the structural rigidity of universities, focusing on research excellence and building an ecosystem for innovation, which became the foundation of NAAMII with three key pillars: Research, Education & Outreach, and Industry & Innovation.

Inspired by INRIA (where I did my PhD), we named our institute Nepal Applied Mathematics and Informatics Institute for Research (NAAMII). We registered NAAMII as a company not-distributing-profits in 2018 so that unlike traditional non-profits, it could sustain itself by reinvesting profits into its mission. By this time, I had also grown increasingly aware of the need for Global South institutions dedicated to solving local challenges and creating knowledge where it’s most needed. My work in top labs during my PhD and Postdoc strengthened my conviction that NAAMII could be a model for such centers beyond Nepal.

After NAAMII’s official registration, we organized the 2018 Annual Nepal AI School, our first major event, which brought together over 100 participants from different countries. The excitement and thirst for knowledge I saw in the young participants left a profound impact on me. Back in London, I couldn’t sleep, questioning why I was waiting to return. With my mind already set on NAAMII, I resigned from my job and moved back to Nepal, fully dedicating myself to building the institute I had dreamed of for so long, and where a close group of friends shared the same vision and strongly supported me to build the institute collectively. Today, NAAMII is a thriving center of excellence, dedicated to fostering AI-driven research, building local talent, and pioneering Nepal’s journey on the global science and technology stage.  

My research group at NAAMII, Transforming Global Health with AI (TOGAI), is advancing AI to tackle critical healthcare issues across the four healthcare areas:

Prevention, Awareness, and Access to Information; Empowering communities with the tools and information to prevent diseases and improve access to healthcare knowledge. Enhancing Clinical Services: Applying AI to improve diagnostics, treatment options, and surgical interventions, making healthcare more effective and accessible.

Health Professional Training and Capacity Building: Developing AI-driven solutions to support the training and skill-building of health professionals, particularly in low-resource settings.

Fundamental Disease Understanding: Conducting in-depth research on diseases to better understand their mechanisms and aid in the development of targeted treatments.

Our work is guided by the specific healthcare needs of low- and middle-income countries (LMICs), where addressing these challenges can have global significance. We approach these issues from four perspectives:

Improving Primary and Community Health Centers: We focus on enhancing healthcare delivery in primary settings by using existing resources like frontline health workers and affordable devices. Through AI-powered task-shifting, we aim to equip healthcare professionals with AI tools that allow them to perform tasks typically requiring specialized expertise, thereby expanding the scope of healthcare services. AI-assisted cervical cancer screening, obstetric scanning and micrsocopy are some of the projects we are doing along this direction.

Developing Affordable and Accessible Technologies: We build solutions tailored for resource-constrained settings, such as our 2D-3D bone reconstruction from biplanar X-rays, which enables surgeons to plan surgeries more effectively when CT scans are unavailable or too costly.

Optimizing High-Burden Tertiary Centers: We aim to ease the workload in tertiary centers, where resources are often stretched thin. For example, our work in scoliosis Cobb angle estimation helps radiologists and surgeons save time by automating measurement processes in high-demand facilities.

Addressing Neglected and High-Burden Diseases: We identify and target high-burden or under-researched diseases where AI can make an impact, such as cervical cancer screening, advanced obstetric scanning, and AI-powered detection of diarrhea-causing parasites. We also conduct foundational research to understand disease mechanisms, such as imaging to study drug resistance in tuberculosis.

Through TOGAI, we aim to develop practical, scalable AI solutions that address the pressing healthcare needs of LMICs, providing meaningful improvements in access, quality, and outcomes in healthcare. Among them the biggest focus now has been in AI-powered task-shifting.  

I had a fruitful stay with Ender who shared the passion for working on a research project that can have a big impact. One of the ideas that we want to explore is around building next generation AI-powered hardware that is low-cost, portable, but can integrate in a single device a wide range of biomedical sensors that are important for clinical diagnositics. My current work in AI-powered task-shifting is tackling multiple diseases by building separate AI models. The next step I was planning and shared with Ender was about doing implementation research on combining multiple AI services to a single person to see how much we can broaden the scope of task-shifted services through a single person which is extremely difficult if it is not assisted by AI or experts. Ender then proposed a great idea on why not we think of broadening the scope at the hardware level as well where we could integrate multi-modal data and have a device that can get information from various sensors such as ultrasound, EEG, ECG, sound etc. And build AI models that can work in smaller GPUs so that we can do offline (on-site) AI. This is the project we will be exploring and planning to apply for funding.

Another collaboration we are looking for is on having a joint PhD student, specifically research associate who is working with me at NAAMII in Nepal but would be enrolled as a PhD student at ETH Zurich (NAAMII cannot provide degree as we are not a degree granting institution). We have the candidate already working at NAAMII and doing really well in an exciting project.

Finally, we would like to explore if there is a possibility of a joint Msc and PhD programs with ETH and NAAMII, focusing on AI and healthcare. The topics we specialize could be around AI and Global health. More about this, you can read in response to the next question

As we aim to make an impact globally, collaborations and partnerships with institutions like ETH Zurich will be crucial. I truly believe that we need more multidisciplinary AI Centers of Excellence (AICEs) in LMICs that can tackle and provide solutions to the big challenges the world is facing. I think these AICEs in LMICs could be a network in itself with we could build a parallel network of partner centers/institutions in High Income Countries (HICs) comprising of big excellence institutions like ETH Zurich. These two can then work together in a coordinated way complementing each member’s strengths so that we can tackle the big challenges together in a more efficient and optimal way, maximizing the limited resources the world has on research.

While the above is a bigger vision and long-term goal, specifically in the present context, institutions like ETH Zurich could really support is through their experience and name on academic excellence and structure. For countries like Nepal, most young people do not want to stay there for higher education. For example, almost every research assistant or associate at NAAMII will go abroad for Msc or PhD abroad because there is no option in Nepal. This is the most important area where ETH Zurich, their faculty and researcher could help, by building a joint graduate program so that the youths have an option to work with us and people at ETH Zurich, so that we have more fantastic students who work for longer term on problems that are important for us to tackle.