zengxiao-he/tessera · AI Repo StatsZEzengxiao-he/tessera
From teacher to tiles — a from-scratch LLM distillation & serving engine: custom Triton/CUDA kernels, FSDP distillation, paged-KV continuous batching, speculative decoding, a Rust gateway, a JAX oracle, and interpretability tooling.
LLM & foundation modelsPython 8created 2026-06-05 #cuda#flash-attention#fsdp#inference-engine#jax#knowledge-distillation#kv-cache#llm#mechanistic-interpretability#ml-systems#paged-attention#pytorch#quantization#rust#speculative-decoding#triton