About

I am an Associate Scientist at USRA working with NASA’s Black Marble Science Team to tailor machine learning algorithms for studying the Earth at Night using satellite observations. The derived inferences have a broad variety of applications ranging from monitoring nighttime clouds, thermal anomalies such as fires and gas flaring to studying urban areas for tracking urbanization, power outages and disaster impact, conflicts, and electricity access to accelerate downstream analysis.

Prior to this I was a NASA Postdoctoral Fellow at Goddard Space Flight Center and Universities Space Research Association, working with the Black Marble Science Team. Previously, I was a doctoral student in Computer Engineering with Profs. Antonia Papandreou-Suppappola and Philip Christensen at Arizona State University focusing on modeling satellite image time-series for change and novelty detection in Earth and Planetary observations.

My research interest lies at the intersection of machine learning, statistical signal processing and remote sensing with applications in Earth and Space Sciences. My projects are primarily centered around unsupervised learning, change, anomaly and novelty detection, time-series analysis. These translate to a variety of research areas in remote sensing such as multitemporal analysis of multispectral Earth and Space Science datasets with applications in detecting, understanding and forecasting changes, natural hazards, extracting rare class signals for scientific discovery, and utilizing satellite observations for monitoring the variability over time. With the growing volume of satellite observations, a major bottleneck for applying machine learning methods is the very limited availability of labeled training datasets. My research is also exploring unsupervised and self-supervised methods for tackling the lack of labeled datasets in remote sensing.

At present, I also volunteer with the NASA SMD AI/ML working group where I actively participate in conducting surveys, hosting workshops, co-convene yearly meetings at the American Geophysical Union Fall Meeting centered around cross-divisional applications of machine learning. I also participated in the yearly group activity for generating training datasets for Earth Science to improve machine learning algorithms. I serve as a technical co-lead in the IEEE Geoscience and Remote Sensing Image Analysis and Data Fusion Working Group.

Broadly, I am interested in applying machine learning advances to remotely sensed Earth and Space datasets and extensions to other scientific applications.

To get in touch please write to srija6791[at]gmail[dot]com, schakr34[at]asu[dot]edu or schakraborty[at]usra[dot]edu.