Beginnings: An Unexpected Course

In 2014, during the first semester of my doctoral program, my former PhD advisor (and now collaborator) suggested that I take a course on ecological dynamics. At the time, I was a first-year PhD student in interdisciplinary engineering at Northeastern University in Boston, where I was getting exposed to complex networks, hydrological simulations, and built systems modeling. Ecology sounded distant and abstract. I remember smiling politely and asking if he was serious. When he nodded, I replied, “Why not?” That small decision had a significant influence on my scientific thinking. 

The course, taught by Professor Tarik Gouhier, introduced a world where stability emerged from interaction and feedback rather than control. Equations described not just flows of water or stress in bridges, but the cooperation and competition that shape living systems. I began to see that resilience was not merely resistance to change, but rather the capacity to adapt. Systems, whether ecological or engineered, survive because they reorganize when disturbed. Those early insights became the seed for a decade of work that would later connect my Machine Intelligence and Resilience (MIR) Lab at IIT Gandhinagar with the Sustainability and Data Sciences (SDS) Lab at Northeastern University.

Early Lessons in Variability

When I joined the SDS Lab, one of the major projects underway was what later became Wang et al. (2015, Nature). Although I was not part of the author team, I was fortunate to attend the lab discussions that shaped the work. The study examined how wind-driven coastal upwelling, which sustains marine productivity by bringing nutrient-rich water to the surface, might shift in a warming climate. The findings revealed that upwelling seasons would begin earlier, last longer, and intensify at higher latitudes, fundamentally altering the rhythm of the oceans. For a young researcher, those conversations were transformative. I realized that climate change was not simply about temperature increasing, but about the rearrangement of time, space, and feedback. Climate systems, like engineered ones, have timing built into their stability. When those rhythms shift, everything connected to them must adapt. That insight, first glimpsed through discussions of winds and currents, became the conceptual thread linking my later work on floods, cities, and networks.

A Pause and a Return

After completing my PhD, I returned to India and established the Machine Intelligence and Resilience Lab at IIT Gandhinagar in 2019. Our focus was on infrastructure resilience, flood modeling, and physics-guided artificial intelligence. Around that time, Kate Duffy et al. (2022, Nature Climate Change) published a study showing that increasing climate variability, rather than mean warming alone, amplifies extinction risks across ecosystems. Although I was not involved in that paper, it struck a deep chord. Reading it felt like rediscovering a conversation I had left unfinished.

That study reminded me that uncertainty is not just a source of error but a fundamental property of complex systems. It re-ignited discussions with Auroop and Tarik, bridging the years since my PhD and setting the stage for a renewed collaboration between MIR and SDS. Together, we began asking how the lessons of variability that shaped ecological understanding could guide the design of resilient infrastructures and climate-ready cities.

Restoring What Collapses

Our first joint study after this reconnection, Bhatia et al. (2023, Communications Biology), focused on how ecosystems recover after collapse. Using thirty real and twenty-seven synthetic plant–pollinator networks, we tested alternative restoration strategies and found that reintroducing the most connected species first, the generalists, leads to nearly optimal recovery. Complex and computationally demanding strategies provided little additional benefit.

The result carried a broader message. Recovery begins by strengthening the key connectors that hold a system together. Simplicity can be powerful when guided by structure. For engineers accustomed to optimization under uncertainty, this felt intuitively right. The work also held personal significance: it was my first publication with Professor Gouhier, the same instructor who had once introduced me to the field of ecological dynamics. What began as a course requirement had evolved into a collaboration, uniting mentor, teacher, and former student through a shared curiosity about resilience.

Managing Before Collapse

If that study explored recovery after disruption, the next collaboration, Datta et al. (2025, Communications Earth and Enviroment), turned to the challenge of prevention. We coupled CMIP6 climate projections with network-based population models across eleven ecosystems to assess how warming alters plant–pollinator interactions. The results were sobering: tropical ecosystems, already near their thermal limits, could lose around half their pollinator populations by the end of the century, while temperate systems remain comparatively buffered.

To operationalize these insights, we introduced the Management Efficiency Ratio (MER), which measures ecological gain per unit of management effort. The analysis showed that multi-plant management strategies yield high returns in the tropics but limited benefits in temperate regions. The larger principle is universal: resilience must be context-specific. Whether managing ecosystems or city infrastructure, equal investment everywhere is neither efficient nor equitable. Understanding where vulnerability concentrates is the first step toward fairness in adaptation.

Toward a Framework for Uncertainty

The collaboration evolved again with Rachindra et al. (2025, npj Climate and Atmospheric Science), a Perspective article that argued for rethinking how science approaches internal climate variability. We proposed that variability, rather than being treated as noise, should be viewed as information about the system’s internal dynamics. The paper outlined how nonlinear dynamical (NLD) approaches — including Lyapunov exponents, attractor reconstruction, complex network analysis, and entropy measures — can be integrated with initial-condition large ensembles to capture the full, irreducible spread of climate outcomes.

The Perspective presented five priorities: connectivity and causality, predictability, change detection, uncertainty quantification, and decision support. It suggested that these methods could help translate the mathematics of complexity into tools for climate adaptation and risk management. In many ways, this paper synthesized the entire arc of the MIR–SDS partnership. If Wang revealed how climate change shifts the timing of natural systems, Duffy reminded us that variability drives risk, Bhatia and Datta demonstrated how ecosystems fail and recover, then Rachindra asked the unifying question: how can we study and manage all such systems together under irreducible uncertainty?

Lessons Across Systems

Across these studies, a single theme endures. Systems, whether ecological, climatic, or urban, thrive on diversity, feedback, and adaptation. They falter when rigidity replaces responsiveness. Resilience is not a static feature to be engineered once; it is a dynamic capability that must evolve. From restoring lost species to designing flood-resistant cities, the essence is the same: to anticipate disruption, reorganize intelligently, and recover function without losing identity. In this sense, the boundaries between ecology and engineering blur. Both disciplines seek to understand networks that can bend without breaking.

Full Circle

Looking back, I see how a single suggestion — to take an ecology course that seemed far outside my domain — shaped an entire philosophy of research. A decade later, I lead a lab that studies how infrastructures, ecosystems, and communities adapt to climate variability, working alongside the advisor who first encouraged me to take that course and the instructor who taught it. Together, through the partnership between MIR and SDS, we continue to connect the dots from coastal upwelling to pollinator networks, from rainfall extremes to urban resilience.

Science, I have learned, resembles the systems it studies. It grows through relationships that span generations and disciplines. Resilience, too, is built on those relationships — between oceans and cities, plants and pollinators, data and decisions, teachers and students. It is the capacity to stay connected, to keep learning, and to adapt gracefully when the world changes. That lesson, first encountered in a classroom discussion on ecological dynamics, continues to guide my thinking, teaching, and collaborations today.