AI Daybook

From the archiveNo. 2

RepoOpen Source & Tools

Tessera, an LLM distillation and serving engine with custom CUDA kernels

zengxiao-he/tessera

GitHubTrending 8

Summary

Tessera is a from-scratch LLM distillation and serving engine featuring custom Triton and CUDA kernels, FSDP distillation, paged-KV continuous batching, speculative decoding, a Rust gateway, and JAX oracle tools.

Read the original

#tessera#llm-distillation#serving#triton#cuda

142 more stories in this issueNo. 2 →Tue, Jun 9, 2026