March e-Newsletter Community Guest Spotlight with 2025 WaterSoftHack Fellows
Posted Mar 10, 2026
Clemson University, in partnership with the University of Iowa, Tulane University, and CUAHSI, is continuing to grow WaterSoftHack, a multi-year NSF-funded initiative designed to cultivate the next generation of water science and engineering researchers. As the program enters its third year, CUAHSI had the opportunity to connect with a few of last year's fellows, including one returning participant and two first-time attendees, to hear about their experiences. Fellows shared what drew them to the workshop, the skills and training they gained, and how they plan to apply these methods in their research. They also reflected on the collaborative opportunities the workshop provided and how it has helped expand their networks within the water science community.
Meet the fellows!
Moses Kiwanuka, PhD Candidate | WaterSoftHack Fellow '24 & '25 | Earth Systems Science | Florida International University
My primary motivation to attend the workshop for a second time was to further strengthen my skills in applying machine learning techniques to water science problems. As a Ph.D. candidate working on water quality monitoring using remote sensing and machine learning, I saw the second workshop as an opportunity to deepen my technical capacity and learn new modeling approaches that could support predictive water science applications, particularly in developing predictive models for environmental variables in large lake systems. I was also motivated to reconnect with the collaborative learning environment and continue engaging with the broader water science community.
The first workshop primarily introduced cyberinfrastructure tools and platforms for water science, including HydroLang for hydrologic data processing and visualization. The second placed greater emphasis on machine learning methods and their applications in water science research. Going in the second time, I had clearer research objectives and more focused expectations around understanding machine learning techniques and exploring how these models can be applied to forecasting environmental variables such as sediment dynamics. That clarity allowed me to engage more actively in discussions and collaborative activities.
The second workshop expanded on the cyberinfrastructure foundations from the first by focusing on machine learning approaches for predictive modeling. These skills are directly relevant to my doctoral research, which uses machine learning and remote sensing to monitor and predict water quality in inland lakes. Together, both workshops have strengthened my ability to work with large environmental datasets, develop predictive models, and interpret results for water management applications.
Through team-based projects and interactive training sessions, I worked with a new group of participants on a project comparing machine learning approaches for sediment forecasting. Working with people from different backgrounds encouraged knowledge sharing and offered new perspectives on machine learning applications in water science. I also came away with an expanded professional network of researchers and practitioners who share similar interests in data-driven environmental modeling.
The hands-on machine learning training and collaborative project work were the most impactful. The opportunity to compare different machine learning models for sediment forecasting was particularly valuable as it demonstrated how different algorithms perform under varying conditions and highlighted the importance of model selection and evaluation. This practical experience deepened my understanding of predictive modeling and its potential applications in environmental monitoring and water resource management.
During the workshop, our team worked on a project titled "Comparing Machine Learning Approaches for Sediment Forecasting," which explored how different machine learning models can be applied to predict sediment dynamics in water systems. Although the work is still being refined, it provided valuable insights into collaborative data science approaches for water science research and strengthened my capacity to apply machine learning methods to environmental forecasting problems.
Aashish Gautam, Graduate Research Assistant | WaterSoftHack Fellow '25 | Jackson State University
I first learned about WaterSoftHack through my research advisor, who encouraged me to apply, highlighting it as a program that would complement my ongoing work in water science. My primary motivation was to strengthen my understanding of data-driven approaches for water resource applications — particularly machine learning techniques that could enhance the remote sensing and hydrological modeling work I was already pursuing as a graduate researcher.
While I had foundational knowledge of machine learning, the workshop gave me much deeper insight into the practical side of deep learning — specifically how to tune hyperparameters, adjust epochs, and make meaningful model architecture decisions. These were gaps I didn't fully realize I had until the workshop filled them, and that knowledge has since informed how I approach model development in my own research.
Through the group project, I had the opportunity to collaborate with three PhD students from three different universities, which made for a rich, interdisciplinary team dynamic. Even though our GRACE downscaling project didn't achieve the target accuracy, the learning experience was invaluable — and so were the connections. Those relationships have meaningfully expanded my professional network within the water science community.
Without a doubt, the group project was the most impactful experience for me. Working through a real problem with peers from different institutions pushed me in ways that lectures alone cannot. Equally memorable was the mentorship from Dr. Samadi, whose genuine willingness to engage with participants and offer guidance made the entire experience feel both supportive and intellectually enriching.
The skills I gained at WaterSoftHack directly influenced my current research. One tangible outcome is a Google Earth Engine application I developed called Sat2SSC, which uses satellite imagery to estimate suspended sediment concentration using a Random Forest model. Behind the scenes, I've also been experimenting with XGBoost and LSTM approaches locally, and I'm actively working to improve the model's accuracy and scalability — work that traces its roots back to what I learned at the workshop.
Hassan Saleh, PhD Candidate | WaterSoftHack Fellow '25 | Geosciences| Western Michigan University
I learned about WaterSoftHack through a LinkedIn post by Dr. Vidya Samadi. The program's agenda and structure immediately appealed to me, especially the combination of both theory and application, and the opportunity for project collaboration, supported by Dr. Samadi and her research group. My primary motivation was to learn advanced machine learning techniques and explore how I could apply them to my own research, making this program a perfect fit for me.
I was particularly interested in learning the theory and application of Long Short-Term Memory (LSTM) models and exploring how to apply them to downscale coarse-resolution satellite data. Because the program was both intensive and comprehensive, I had the opportunity to build a model from scratch, understand its underlying structure, and learn how to tune hyperparameters both theoretically and practically. During the second week, several colleagues showed interest in the downscaling project, and we collaborated to build and apply an LSTM model over Arizona to downscale GRACE data. Although the model ultimately required further tuning to optimize results, the workshop provided a kickstart in building and implementing advanced models that I hadn't had access to before.
One of the most rewarding aspects of this program was building connections with other graduate students from diverse backgrounds. This collaborative environment, which the program provided, is something I also valued during my time at the CUAHSI Water Prediction Innovators Summer Institute at the University of Alabama in 2024. Our cohort continued to network beyond the workshop; we connected on LinkedIn, learned about each other's research, and even had the chance to meet in person during the AGU Fall Meeting.
For me, the most impactful part of the workshop was the pairing of theory with immediate application. I also enjoyed the collaborative environment, which made the problem-solving process easier. We regularly discussed challenges and worked to find solutions in real time. I was thoroughly impressed by how professional and effective the communication was, even in a virtual setting, which helped our project progress smoothly despite the challenges we faced.
As a result of the workshop, our group drafted a short paper that we submitted to the coordinators, and we received valuable feedback in return. While the code and initial results require more time and effort to be fully organized for public sharing, the learning experience in developing these models was immense and directly beneficial to my ongoing work.