Anwer
أنورAn engine-agnostic semantic profiler for LLM inference systems.
Works across vLLM, SGLang, TensorRT-LLM, and NVIDIA Dynamo — a single layer for understanding where time and memory actually go during inference, and why.
What "semantic" means here
"Semantic" refers to inference-execution semantics — scheduler decisions, KV cache behavior, expert routing, prefill/decode phase boundaries — not text or content semantics. Anwer joins that semantic control-plane with the hardware data-plane (kernel timelines, CUPTI / torch.profiler traces) on a shared timeline, so a latency spike can be attributed to what the engine was doing, not just which kernel ran.