Sneha Muppalla

smuppalla06[at]berkeley[dot]edu

Hello Hello! I am an undergraduate student at UC Berkeley studying Electrical Engineering & Computer Sciences (EECS) and Business Administration under the M.E.T. program.

I work on low-level optimization and computer vision architectures for large-scale physical models. I also write and direct films.

Active Work

HiPeRLab (UC Berkeley) — Undergraduate Researcher
Developing JAX pipelines and differential optimization frameworks to compute backward-pass gradients and map trajectories through continuous physical boundaries.
RunAnywhere — Engineering Intern
Inference layer: Metal kernels for Apple Silicon, Qualcomm Hexagon NPU, KV-cache optimization, and the Python eval harness.
CS 61C Course Staff (UC Berkeley) — Teaching Assistant
Teaching computer architecture fundamentals, focusing on CPU pipeline logic, caches, memory-mapping , and data-level parallelism.

Projects & Research

Stormscope — github
An extreme weather nowcasting project using physics-guided sampling networks to track atmospheric anomalies across high-dimensional climate meshes. Built to scale via containerized cloud environments.
VGGT Inference Optimization — github
Profiled and optimized inference for VGGT (CVPR 2025 Best Paper) using Nsight Systems. Refactored sequential frame attention to fused batch processing, implemented early layer global → frame attention pruning, and enforced FlashAttention-2 via SDPA with memory contiguity.
Integrating Audio-Visual Features for Multimodal Deepfake Detection
Sneha Muppalla, Shan Jia, Siwei Lyu
IEEE MIT URTC, 2023

Interests

Systems & Infrastructure: Low-level hardware compilation, inference engine efficiency, distributed training optimization, and automatic differentiation frameworks.
Applied Modalities: Computational earth systems, atmospheric dynamics, planetary data arrays, and accelerated material discovery. Hobbies: Narrative filmmaking (screenwriting and directing), climbing, basketball.