Chop Wood & Carry Water

Linux · Networks · Wireless · AI · IT
“Before enlightenment, chop wood and carry water. After enlightenment, chop wood and carry water.”
— Zen proverb

My daily work is a "meditation" on IT infrastructure: I configure Linux servers, secure networks, and build AI servers. The tools may change, but the mindful discipline remains the same.

⏳ Professional Journey

2011 – Certified Network Deployment & Integration Engineer
Earned certification in communication systems — wireless systems configurations, troubleshooting, O&M, and live demonstrations. This laid the groundwork for years of field engineering across the world. Radio transmitting licensed in multiple countries.
2014 – Present · IT Infrastructure & Systems Administrator
Transitioned into a full Linux/network administration role, eventually specializing in Life Science IT.

⚙️ Core Competencies

🐧 Linux & High Availability

  • Red Hat Enterprise & Ubuntu Server
  • The Full Spectrum: System design, integration, deployment
  • Performance tuning & troubleshooting
  • Clustering & failover configurations

🌐 Networking & Security

  • Next‑Gen Firewalls (Palo Alto, Sophos UTM/XG)
  • DNS, DHCP, VPN, VLAN, Routing, NAT
  • Cyber Security audits & webserver hardening
  • Enterprise CISCO Wireless Networks
  • CrowdStrike, Splunk, Cloudflare

☁️ Cloud & Virtualization

  • Amazon AWS (deployments, migrations)
  • On‑premise & hybrid architectures
  • Hypervisors: VMware ESXi, Proxmox, RedHat Virtualization
  • Containers: Docker, Podman, LXC

🤖 AI / ML & HPC

  • NVIDIA Enterprise AI & Deep Learning servers
  • Big Data AI/ML projects
  • Dell & Supermicro compute & storage
  • Maintainer of the company's AI infrastructure

🔐 Compliance & Data Privacy

  • GDPR & Highly confidential data handling
  • Privacy-enhancing computational techniques
  • Confidential computing
  • Cryptographic policies & audit trails
  • Secure multi‑tenant environments
  • NDA, Security Clearance

🛠️ DevOps & Collaboration Tools

  • Atlassian: Jira, Confluence, Bitbucket
  • GitLab, Subversion, Nexus Repositories
  • CI/CD pipelines & automation scripts
  • AWS CloudFormation

🧠 The AI & Machine Learning Arc

My journey into artificial intelligence began well before the current hype. Around 2017–2018, I started working with computer vision and machine learning at my workplace, when the field was still in its infancy. By 2023, I had added large language models — the technology most people now simply call “AI” — to my toolkit. What started as a professional requirement quickly became a genuine passion.

Today, I build and maintain both computer vision systems and massive language models, both professionally and personally. The engineering discipline required to train, optimize, and deploy these models is immense — but so is the reward. Seeing a model recognise objects, understand context, or generate creative text from a handful of GPUs is nothing short of magical.

“AI is not just a tool; it’s the natural extension of a curious mind. The deeper I go, the more I realise there is to learn.”

🤔 Will AI Replace Your Job?

Software Engineer working at McDonald's (meme)

“Will AI – Artificial Intelligence replace your job?” — that’s a question many people are asking, and not just in the software industry.

The answer is both yes and no.

Historically, steam engines gave way to diesel, diesel to electric, and wheels to maglev (magnetic levitation). In the same way, the traditional way of working will indeed be replaced. But that doesn’t mean humans become obsolete.

Why “no”? Because humans will step into new roles — as Project Managers or Technical Account Managers (TAMs) — overseeing every step with full contextual understanding, without needing to read every single line of code. However, the technical skill must remain: engineers must be able to read, interpret, and cross‑check what the AI has produced, and understand the “why” behind each decision. In short, real, skilled engineers will still be essential — just with less “chopping wood” and more strategic oversight.

The tools change, but the need for human judgment, creativity, and accountability endures.

🤖 The New Way of Software Engineering

The practice of software engineering is undergoing a profound transformation. What was once a solitary pursuit — the lone engineer wrestling with code, tests, and elusive bugs — has evolved into a collaborative partnership between human expertise and artificial intelligence.

“With AI as a collaborator, we can tackle software engineering challenges that would otherwise feel insurmountable — projects so complex and ambitious that we would never even consider starting them alone.”
— Zsolt Tari
⛓️ The Old Way
  • Coding & Configuration — a solitary endeavour, often in isolation.
  • Testing — manual or semi‑automated, slow, and prone to oversight.
  • Troubleshooting — an agonisingly slow process of elimination, consuming hours or even days.
  • Team dynamic — the engineer carries the weight alone, with little support from automated intelligence.
One person, one stack, one struggle.
The New Way with AI
  • Coding & Configuration — AI agents, equipped with full tool access, translate written prompts into working code and configurations with remarkable speed and precision.
  • Testing — AI agents execute tests autonomously, identifying failures and discrepancies without human intervention.
  • Troubleshooting — AI operates at a speed that surpasses human capability: it reads multiple documents simultaneously, monitors logs in real time, and executes parallel diagnostic tasks while continuing to compile and configure.
  • Human Guidance — the engineer remains essential, providing strategic direction, understanding the broader context, and making critical decisions about what to build and why.
  • Guardrails & Approval — AI agents inevitably encounter guardrails: security boundaries, ethical constraints, and policy limitations. These are intentional, and human approval is required to proceed, ensuring safety and accountability.
  • True Collaboration — this is not automation replacing the engineer; it is augmentation. The human feels a sense of teamwork, not isolation. Problems are tackled together, not alone.
Human + AI = a partnership that multiplies capability.
🤝
From isolation to collaboration. The old model was a lonely grind — the new model is teamwork. The AI does the heavy lifting of implementation and debugging, while the human provides vision, oversight, and the irreplaceable spark of creativity. No more suffering alone — it’s a partnership where both sides bring their unique strengths to the table.
This isn’t science fiction. It’s the everyday reality of modern software engineering — and it’s only the beginning.

🌍 Global Impact in Information Technology

As a multi‑customer engineering team leader, I’ve led system implementations from design to integration — on premise, in the cloud, and hybrid. My work has touched 17 countries, supporting leading mobile telecom operators and, later, a global biopharmaceutical software company.

At work, I handle sensitive data, build and secure server environments, and maintain the backend for enterprise software that accelerates production. I regularly conduct cyber security tasks and optimize server protections to keep our platforms compliant and robust.

“I’ve configured servers in many different data centers, troubleshot networks around the world, and built Artificial Intelligence servers — but always returning to the same command line, the same mindfulness.”

🏠 Home Lab & Off‑Grid AI

Technology doesn’t stop at the office door. My home lab is a playground for high‑performance computing, where I experiment with virtualization clusters, container orchestration, and smart home devices. IoT sensors, automated lighting, and energy monitoring keep the physical world responsive and efficient.

But the crown jewel is the AI server. I run real‑time computer vision inference and host a 72‑billion‑parameter large language model — a system of staggering versatility. It is fluent in 29 human languages, with comprehension support for 119 languages. It writes code across over 90 programming languages, excels at analyzing charts, patterns, and layouts, and understands images and videos with remarkable depth. All of this runs entirely offline, without ever phoning home to a cloud provider. It's the ultimate privacy‑respecting research assistant: always available, never connected, and entirely under my control.

“The future is already here. It speaks 29 languages, writes code in 90, and sees the world through any lens — all powered by a handful of GPUs in a custom‑built server rack, with a cooling solution I designed and 3D‑printed at home.”

Zsolt's Self Hosted Artificial Intelligence Architecture

XMPP Chat Server Self‑hosted · Encrypted

Conversations (Android) Gajim (Linux)
End‑to‑End Encryption · OMEMO
RAM: 2‑4 GB Disk: ~500 MB

Telegram Backup Channel

Fallback Instant
Bi‑directional relay
RAM: ~512 MB Internet required

Hermes Desktop App Cross‑platform

Linux Windows
Laptops & PCs via encrypted channel
End‑to‑End Encrypted Real‑time sync
|

Hermes Agent Core Orchestrator · Tool‑Calling

Multi‑model router KV cache optimized Persistent context
Routes: STT → LLM → TTS / SD / Vision
RAM: 4‑8 GB CPU: 4+ cores
|

Vision AI · Qwen3.8‑27B August 2026 Release · Thinking Model

Vision capable Reasoning KV cache optimized
Fine‑tuned for Hermes
VRAM: 22 GB RAM: 32 GB+ Disk: ~18 GB

Image Generation · Stable Diffusion

SD 1.5 (512x512) SDXL (1024x1024) SDXL Turbo
Refiner + Turbo for enhancement
SD 1.5: 4 GB SDXL: 8 GB SDXL+Refiner: 12 GB RAM: 16 GB Disk: ~10 GB

Speech‑to‑Text · faster‑whisper

90+ languages Transcribing from voice or video large‑v3 FP16 optimized
Handles accents & noise
VRAM: 4‑6 GB RAM: 16 GB Disk: 2.9 GB

Text‑to‑Speech · NeuTTS Air

Voice cloning 24 kHz PCM Reference: jo.wav
Full FP16 · maximum quality
VRAM: 6 GB RAM: 16 GB Disk: ~2 GB

💎 Hardware cost estimation for such system based on 2026 prices:

Budget Tier
Estimated Total Cost
Minimum (Good Value)
~$10,000 – $12,000
Typical (Balanced)
~$13,000 – $16,000
Premium (Max Performance)
~$17,000 – $19,000+
Infrastructure beyond the core hardware: For true resilience and self‑sufficiency, I strongly recommend complementing this system with offsite storage (co‑location) for secure backups, a solar photovoltaic energy supply, and uninterruptible LFP battery backup power solutions. These ensure continuity, energy independence, and peace of mind — and I have personally implemented all of them. (These are not included in the budget estimation above.)