
Ember
Adaptive local LLM inference for macOS.
What I enjoy working on.

Find, fit, and benchmark local AI models on your Mac.

Implemented the FA2 paper in CUDA for efficient attention.

Medical transcription and care plans at 4.14% word error.

Detect Ethereum fraud live with 99% precision.

CUDA matmul at 90.94% of cuBLAS, plus parallel reductions.

Higher Solana staking rewards for greener purchases.