raw.sh

R

robert washbourne

@rawsh0

posttraining lead @ zyphra

interests:

•

synthetic pretraining

•

efficient reasoning

•

ttc

Preprints

ZAYA1-8B Technical Report, 2026

[Paper], [Model], [Project Website]

Training Foundation Models on a Full-Stack AMD Platform, 2025

[Paper], [Model]

posts

Batched reward model inference and Best-of-N sampling - 11/19/2024

Teaching chat models to solve chess puzzles - 8/24/2024