This is your work, valued

San Francisco, CA

Isaac Rehg

Intermediate
@IsaacRe

Founding engineer working on lightning-fast LLM inference.

vllm-kvcompress. KV cache compression for high-throughput LLM inference

159

Syntactically-Constrained-Sampling. LLM sampling method for enforcing syntax adherence in generated output

25

tacklebox. Improved handling of PyTorch module hooks

5

weighed-levenshtein-substring. Fork of https://github.com/infoscout/weighted-levenshtein

2

splatnet. Python

1