Benchmark LLM Serving Without Lying to Yourself
Understand inference performance as a workload-dependent surface, then test saturation and goodput with a controlled local benchmark.
Production AI engineering: reliable agents, efficient inference, and research implemented and tested.
I'm a software engineer focused on how AI systems behave beyond the demo. I write about agent execution, inference performance, and the engineering trade-offs behind research ideas.
Local inference performance, reliable agent execution, and practical implementations of research papers.
Understand inference performance as a workload-dependent surface, then test saturation and goodput with a controlled local benchmark.
A local benchmark harness for measuring latency, throughput, failures, and goodput.