RadixArk builds large-scale training and inference systems, with the aim of making frontier-level AI infrastructure open and accessible.
The work is systems-heavy: distributed training, serving throughput, and the software layer between models and accelerators.
