Carl Pearson is an LTE in the Scalable Algorithms group at Sandia National Labs and a Research Assistant Professor (LAT) of Computer Science at University of New Mexico. He works on future architectures for scientific computing, GPU communication for distributed linear algebra, and GPU acceleration of irregular operations.

He received his Ph.D in Electrical and Computer Engineering from the University of Illinois, and his B.S. in Engineering from Harvey Mudd College.

Running Vicuna-13B in Google Cloud

How to experiment with hosting Vicuna-13B on a cloud VM

Transcribing Videos with OpenAI's Whisper and ChatGPT

Transcribing Videos with OpenAI’s Whisper and ChatGPT (and yt-dlp and ffmpeg and pydub and imagehash and pandoc)

Interesting Links

Here are some interesting things I’ve read. Presented without comment or endorsement.

Composable Systems Opportunities and Challenges

I’m a panelist on the “Composable Systems Opportunities and Challenges” panel at SC'22, with my colleagues Christian Pinto and Kaotar El Maghraoui from IBM Research, the CEO and founder of H3 systems Brian Pan and my Ph.

Making Twitter Tolerable

I’m not much of a Twitter afficionado, which is probably why I prefer to use the following search string:

Implementation of a Integer Linear Programming Solver in Python

Roll your own integer linear programming (ILP) solver using Scipy.

My Favorite g++ Flags for a New Project

C++ is hard; make it a bit easier.

Design of

Design of a static site with GNU Makefiles and pandoc

SuiteSparse Collection Downloader

One of the resources I use to evaluate sparse-matrix vector multiplication (SpMV) performance at Sandia National Labs is the SuiteSparse Collection.

Setting up Python with Pyenv and Poetry on Debian

Setting up a python development environment is annoying. We’d especially like to avoid: