Adithya Reddy Nallabolu of Meta

3D Reconstruction is rapidly evolving—from classical voxel-based methods to neural representations such as NeRFs and Gaussian splatting. While these approaches demonstrate impressive quality, deploying them on AR/VR devices introduces strict constraints around power, memory, latency, and thermal limits.

In this talk, I will explore the current landscape of 3DR techniques and critically analyze their feasibility for real-time, on-device AR/VR systems. Drawing from practical experience building power-efficient 3DR pipelines, I will discuss trade-offs across representations, data movement, and compute efficiency. The talk will highlight key system design principles—such as selective computation, adaptive resolution, and incremental updates—that are essential to bridge the gap between research innovation and product deployment.

Biography:

Adithya Reddy Nallabolu is a Research Engineer at Meta working on EMG-based input for AR glasses, exploring novel human–computer interaction methods for next-generation wearable devices. Previously, at Qualcomm, he worked on 3D reconstruction and spatial computing for AR/VR, focusing on building power-efficient, real-time systems through algorithm–hardware co-design. He has contributed to multiple patents and innovations that improve the scalability and efficiency of on-device 3D reconstruction. 

Address:California, United States