The performance layer behind private AI inference.
MoonMath.ai builds the most efficient, privacy-centric AI inference endpoint.
We are a small team of mathematicians and engineers building fast, private AI inference through low-level algorithms, systems engineering, and hardware-aware optimization.
PostsAll posts
Open sourceUpdated Oct 13, 2026
Zro ♥ open source
A live journal of Zro’s upstream open-source contributions.
Explore our contributions →ResearchJun 23, 2026
HyperQuant: A Rate-Distortion-Optimal Quantization Pipeline
Read post →AMDJun 17, 2026
A Fast Attention Kernel for MI300X, Written in HIP, Not Assembly
Read post →TPUSep 10, 2026
Making GLM-5.2 Sparse on TPUs
Read post →Compression, kernels, and hardware sit under the API.
Agent layer · Works with
Claude Code, Codex, Cursor, Cline, Opencode, Hermes
API layer
Completions and Anthropic Messages
Privacy layer
Multi-region inference with zero request retention
Kernel layer
Custom attention and serving kernels for model-specific throughput
Hardware layer · Runs on
AMD, NVIDIA, and Google TPU deployment paths