platoseedplatoseed
AI that makes AI fast
Wafer builds AI agents that work as autonomous performance engineers, optimizing GPU kernels for AI inference. Our product is serverless and dedicated inference for the world’s fastest open source LLMs, achieved by Wafer's autonomous performance engineers.
prev @ two sigma, google, sei labs, and axlab. cs + econ at uchicago.
AI agent that turns your slow PyTorch into fast GPU code, automatically
Wafer automatically optimizes PyTorch code into GPU kernels, enabling 1.5–5x faster performance and hardware portability across NVIDIA and AMD GPUs. It provides production monitoring and automatic fixes without CUDA expertise, targeting ML engineers and teams with underutilized GPUs or high GPU costs, and invites participation in a pilot program.
Formerly “Herdora” · why startups rename →
AI for AI Infrastructure
Making AI run fast on any hardware.
