Local LLMs & inference
Hands-on with Llama.cpp, Ollama and vLLM — from local prototyping to private deployment on company infrastructure.
Full CV
Full CV
A complete overview — qualifications, tools, methods, and operational discipline.
State Education
Ing. (Master of Science) in Software Engineering
University of Tomas Bata, Czech Republic
Local AI
Data privacy is non-negotiable. I run almost all AI tools locally, ensuring minimum data leaves my machine and full transparency over data collection. I specialise in open-weight / open-source models for maximum control.
RAG architectures
Designed retrieval-augmented generation pipelines and semantic analysis. LangChain, LangGraph, MCP servers.
Multimodal generation
ComfyUI for image, video and audio. I track the best-performing models across modalities.
Coding harnesses
Daily use of OpenCode, Pi and Hermes AI coding harnesses for faster, more reliable development.
Voice & speech
Text-to-speech and speech-to-text models, integrated into software products where they make sense.
OCR & document extraction
Built OCR systems and information extraction from documents — turning scans and PDFs into structured, searchable data.
- Llama.cpp
- Ollama
- vLLM
- MCP servers
- LangChain
- LangGraph
- RAG
- ComfyUI
- OpenCode
- Pi
- Hermes
I continuously stay informed on the best-performing models for image, video and audio generation, OCR, information extraction from documents, RAG, TTS and STT. I can deploy models privately for the company and implement them into software products. Linux is my main OS and is practically always on the deployment servers.
Known programming languages
AI · automation
Python · Julia
LLM/AI integration & automation
Web
HTML · CSS · PHP
Web
Data
SQL
SQLite, MySQL, Microsoft SQL Server
Systems
C · C# · C++
Systems programming
Office
Office & Documentation
MS Office & LibreOffice — advanced Excel filtering, macros and functions
Official certificates
Cisco Certified Network Associate (CCNA 1–3)
Networking fundamentals, routing, switching, VLANs, and security protocols.
Cambridge Assessment English — C1 Advanced
Professional proficiency for technical and business communication.
Online courses
- FreeCodeCamp.org — Responsive Web Design Certification
- Alison.com — Computer Networks Fundamentals
- Alison.com — Cryptography & Data Security
- Alison.com — Time Management & Productivity
Linux
Linux has been my daily desktop and development environment for many years. Comfortable with Bash scripting, terminal workflows, package management, and system-level configuration. Experienced with Windows environments, but I prefer Linux for its transparency, customisation, and security model.
Operational discipline
All work data is encrypted via VeraCrypt (post-quantum encryption) with regular encrypted backups. Data is treated as a first-class security concern — never an afterthought.
OpenBSD
OpenBSD is an option I can offer where a security-first environment matters most. I value its proven security-first architecture, strict code auditing, minimal default attack surface, and out-of-the-box secure networking — and can propose it as a solid foundation for high-security servers and workstations.
Why OpenBSD
Security by design
Proactive mitigations, memory protection and continuous code auditing as a core philosophy.
Minimal attack surface
A lean default install — only what is needed, nothing more to secure.
Secure networking out of the box
Integrated cryptography and a hardened network stack with sane defaults.
Hardware & networking
Custom PC design
Component selection and responsible system builds for development, AI workloads, or general use.
Cisco networking
Configuration of switches and routers, network topology analysis, performance testing, and security audits.
Cross-platform setup
OS configuration and troubleshooting across Windows and Linux environments.
Emacs + vim
Highly proficient in Vim keybindings and terminal-based navigation. Extensive experience using Emacs as a customisable text editor, code environment, and productivity hub with custom local AI integration. Comfortable working entirely without a mouse, leveraging keyboard-driven workflows for coding, note-taking, and system management.
Advanced productivity system
I apply structured productivity methodologies to maintain efficiency across development, security, and AI projects and utilize a highly advanced time-management system integrated into my workflow.
Complemented by formal certification in Time Management.
- Emacs org-mode
- org-agenda
- capture
- time-blocking
- retrospectives
Week-agenda (W47)
Mon 18 Nov
Monday — 18 November
09:00 Deep work — RAG pipeline debug
DEEP
11:30 Code review — security audit
REVIEW
14:00 Model evaluation (vLLM bench)
AI
16:30 Backup verify + log review
OPS
Tuesday — 19 November
09:00 Deep work — Emacs config overhaul
DEEP
13:00 Client call — AI deployment scoping
CALL
17:00 Weekly retrospective
PLAN
Sample Emacs org-agenda week view
Graphics & multimedia
- GIMP & Inkscape — open-source alternatives to Photoshop & Illustrator
- Blender — 3D modelling, animation, and rendering
- Darktable — professional RAW photo processing pipeline
- Adobe Suite — Photoshop, Lightroom, Illustrator, InDesign
Time management
A structured methodology keeps deep work, security review, and AI experimentation all moving in parallel — without burnout.
Method
Time-blocking, agenda-driven planning, weekly retrospectives, and a strict separation between deep work and shallow tasks. Everything is captured in Emacs org-mode and reviewed daily.
Tools
Emacs org-agenda, calendar sync, and a personal reporting dashboard. No cloud dependency.
Time-blocked week
Daily review · 09:00
Deep work blocks
09:00 — 12:00
DEEP
14:00 — 16:30
AI
Shallow / admin
12:00 — 13:00
ADMIN
17:00 — 17:30
PLAN
Sample week — split between deep work, review, and AI experimentation