Los Alamos National Laboratory

Parallel to my academic work, my collaborations with Los Alamos National Laboratory have been pivotal in preparing me to tackle real-world challenges, explore numerous datasets, and come up with solutions independently while contributing to the sector of national security. During my first summer with the Data Science at Scale team, I took charge of developing an end-to-end generative modeling pipeline using Generative Adversarial Networks (GANs) to forecast wildfire simulations from existing scientific data. This project aimed to bring AI/ML workflows to staff scientists and physicists, enabling them to better train, model, and evaluate the vast amounts of simulation and experimental data generated across national laboratories. I presented this work at the Los Alamos National Laboratory Annual Student Symposium with over 500 submissions. Inspired by this experience, I applied to the Los Alamos Dynamics Summer School, a highly competitive fellowship program for interdisciplinary research of dynamic systems. My research there focused on vision-based state estimation for a serial robotic manipulator, where I investigated how spatially structured data representations could enhance visual understanding in robotic systems. I designed a multi-view data collection framework that captured the manipulator from three orthogonal perspectives and integrated them into a unified representation preserving geometric dependencies across views. Using a ResNet-18 backbone as a feature extractor paired with a soft multilayer perceptron (MLP) head, the model inferred 3D joint angles directly from 2D visual inputs without any encoder feedback. This approach enabled robust estimation even under partial occlusion of the end effector. Through simulation-to-real training, the model generalized effectively to physical hardware while minimizing dependence on costly real-world data collection. This work has been accepted for presentation at the International Modal Analysis Conference (IMAC) 2026.

Summary and Reflection

Overall, Los Alamos has contributed to my journey as a researcher in many ways. Through GML, I knew my work had to be centered around machine vision and computer vision. Through my experience at Los Alamos, I was able to narrow that further and sharpen my research questions around robotics, digital twin frameworks, and applications in defense and manufacturing. Working independently on high-stakes problems, presenting at a national symposium, and earning a competitive fellowship gave me confidence in my ability to contribute meaningfully beyond an academic setting.

The connection to joy of living here remains the same. The work I have done here is what I can contribute to problems such as wildfire rescue or disaster management through robotics. I see my work being used and valued in scenarios across the world and that is what I truly think brings out the aspect of joy of living for me.