Stanislav Cherkasov

{DevOps,DevSecOps,Platform} Engineer

Senior {DevOps,DevSecOps,Platform} Engineer contracting with US/EU teams. I design, build, and operate cloud infrastructure, Kubernetes platforms, CI/CD, automation, and production systems.

This site collects non-confidential notes from my work and lab practice: Kubernetes, GitOps, Terraform, Ansible, Linux, Proxmox, CI/CD, and security automation.

Personal blog and personal views; always validate commands in your own environment.

Explore all posts

Recent

Speech-to-Text on Intel Iris Xe, Part 4: Open WebUI Dictation, Voice Calls and Long Recordings

16 October 2026·3543 words·17 mins

How I use self-hosted Whisper for Open WebUI dictation, voice calls and long recordings, what everyday use broke, what I changed and what is still unexplained.

Speech-to-Text on Intel Iris Xe, Part 3: Testing Whisper Accuracy with FLEURS and Common Voice

15 October 2026·Updated: 16 October 2026·4577 words·22 mins

How I test Whisper speech-to-text accuracy in four languages with FLEURS, Common Voice and word error rate (WER), and why my gates rejected all three changes.

Speech-to-Text on Intel Iris Xe, Part 2: Building and Releasing Container Images

14 October 2026·Updated: 16 October 2026·4253 words·21 mins

How I build and release Intel GPU speech-to-text containers with OpenVINO, GitLab CI and Harbor, then verify the tested digest is running in Kubernetes.

Speech-to-Text on Intel Iris Xe, Part 1: Giving Idle GPUs a Job

13 October 2026·Updated: 16 October 2026·1959 words·10 mins

I moved Whisper from a CPU-hungry service to idle Intel iGPUs. OpenVINO made it useful; passthrough, warmup, queueing and routing made it operable; repaired speed comparisons showed the real gain.

RTX PRO 6000 Blackwell Max-Q: Proxmox Passthrough for Local AI

7 October 2026·Updated: 8 October 2026·6437 words·31 mins

Running a 96 GB RTX PRO 6000 Blackwell Max-Q on Proxmox and K3s: PCIe topology, VFIO, NVIDIA drivers, vLLM and real onboarding failures.

↑