dan@terminal:~$

cat about.txt

I'm an ML engineer at Unsloth working on open source software for LLM training and fine-tuning. My focus is on memory / speed optimizations and distributed training.

I was formerly an ML engineer at Axolotl AI, implementing various memory and speed optimizations for language model training, scaling it to multi-GPU / multi-node workloads, and generally improving the reliability / usability of our open source software.

Before that, I was at AWS as an ML engineer, where I built object detection for landmines, stood up a few LLM + RAG apps, and did various other applied ML activities.

Even earlier, I was at Sensor Tower modeling active users and retention curves for mobile apps, and at Fomoro AI as a Machine Learning Engineer developing RL models for combine harvester control.

I studied computer science and math at UMass Amherst. During my master's degree, I published some papers at the intersection of ML and comp neuro, focusing on spiking neural networks (SNNs), and built BindsNET.

I have an Erdős number of 2 (path: Paul Erdős → Miklós Ruszinkó → me).

Feel free to send me an email if you'd like to connect!

Dan Saunders

I live in rural Massachusetts with my six sheep, four chickens, and a mini australian shepherd :)

ls -la /education/

M.S. Computer Science

University of Massachusetts, Amherst

Focus: Machine Learning, AI

B.S. Computer Science

University of Massachusetts, Amherst

Focus: Theory of Computation

B.S. Mathematics

University of Massachusetts, Amherst

Focus: Mathematical Computing