← Articles
MH

Marek Horak

Author

DevOps and SRE at a Brno SaaS company, designing on-call rotation systems and incident runbooks. Reduced mean time to resolution by 50% by standardising the incident command structure across 8 squads.

devopsarchitecturetestingperformance

Articles by Marek Horak(3)

Async LLM pipelines with queues and workers

LLM API calls are slow and rate-limited, making synchronous handling a bottleneck for any non-trivial workload. This guide covers designing async pipelines with queues and workers to handle high-throughput LLM workloads reliably in production.

llmbackendarchitectureperformance

19 July 2026

Marek Horak — ANN Tech — ANN Tech