Quantization Through the Lens of CS 61C
An illustrated introduction to numerical formats, integer quantization, memory movement, and why smaller models are not automatically faster.
Read the article →Hi, I’m Sneha.
I study EECS and Business at UC Berkeley through the M.E.T. program.
I’m interested in efficient machine-learning systems, especially how numerical representations, memory, software, and hardware interact. I also enjoy writing, chocolate, and learning filmmaking.
Selected work
On-device ML inference — RunAnywhere
Worked on deploying and evaluating language and vision models on Apple Silicon and Qualcomm accelerators.
Asynchronous function calling — Sky Computing Lab
Investigating latency and context-processing overhead in synchronous and asynchronous tool-using language models.
Cooperative aerial transportation — HiPeRLab
Working on optimization methods for multi-drone payload transportation.
Writing
Elsewhere
When I’m away from my laptop, I’m usually climbing, watching films, writing, or trying to understand how movies are made.
Films, screenplays, and film notes →