-
Open role
GPU Systems Engineer
Posted 1 month ago
We compress LLM KV cache and validate performance (e.g., quantization × prefix dedup, decomposition) across model families (Mistral/Qwen/Llama) with zero re-tuning. First role focuses on the fused Triton/CUDA dequant + attention kernel, turning validated memo…
Location
Remote (US)
Work setup
full-time · Remote (US)
Compensation
Equity-heavy; paid 1–2 week trial.
Benefits
Equity-heavy, Paid 1–2 week trial