Multi-tenant Kubernetes operator for self-hosted GitHub Actions runners. Scale-to-zero workers, per-tenant egress IP pools, and GPU priority scheduling across a shared ResourceQuota — an Actions Runner Controller (ARC) alternative.
Quantize an LLM and check it still refuses what it should — a GPU-aware quantization CLI that measures the safety drift of the quantization it just performed.
Classifies GPUs based on their 3D rendering benchmark score allowing the developer to provide sensible default settings for graphically intensive applications.
LoRA post-training for open-weight models: SFT, GRPO, and on-policy distillation. Describe a run in TOML; Flash allocates a GPU, trains, streams checkpoints, and serves the adapter.