Yash Ranjith
Westmont High School
Campbell, California
Modeling Pollution Spread with Obstructions Using Physics-Informed Neural Networks
Yash found a way to improve outdoor pollution models by designing neural networks that adhere to the laws of physics.
View Poster
Yash Ranjith, 18, of San Jose, created a neural network to speed up the prediction of how pollution spreads for his Regeneron Science Talent Search environmental science project. Environmental disasters can release pollutants that move with wind and water currents, putting local life at risk. Emergency plans like evacuations rely on computational methods to predict how pollution will move. Current models can take from hours to days to calculate these predictions, especially over wide areas and long time periods.
In his project, Yash designed a physics-informed neural network to generate predictions based on partial differential equations. It could model the spread of pollution around obstacles over 2,520 times faster than traditional methods, with minimal cost to accuracy. What would have taken several days can now be completed in seconds with his network.
Yash, the child of Ranjith Thiagarajan and Jayasree Mahadevan, attends Westmont High School. He is an intern at the NASA Ames Research Center researching reinforcement learning to prevent collisions in crowded airspace. In his spare time, Yash writes on Medium about AI applications in climate sciences and has more than 6,500 followers. He also enjoys outdoor bouldering.
Beyond the Project
Yash founded and leads his high school’s competitive programming club. He also founded the sustAInable club, which uses datasets to model environmental threats.
FUN FACTS: In his free time, Yash tends to his small orchard of hybrid fruit trees. Their grafted branches have produced multiple varieties of persimmons, apples and peaches.
