Posts
All posts from William Zujkowski - security insights, AI/ML projects, and career development
Local LLM Deployment: Privacy-First Approach
Complete guide to running LLMs locally for privacy: hardware requirements, model selection, optimization techniques, and practical deployment strategies
Fine-Tuning LLMs in the Homelab: A Practical Guide
Complete guide to fine-tuning open-source LLMs on homelab hardware using QLoRA, covering dataset prep, training optimization, and evaluation
Building a Security-Focused Homelab: My Journey
How I built a comprehensive security lab at home for learning and experimentation—covering Proxmox, VLANs, IDS/IPS, and keeping my family's data safe
Securing Your Personal AI/ML Experiments: A Practical Guide
Lessons from running LLMs and AI experiments at home while keeping data secure, covering model isolation, network segmentation, and privacy controls
From IT Support to Senior InfoSec Engineer: My 15+ Year Journey
The winding path from fixing printers to securing federal systems over 10 years—lessons learned, mistakes made, and advice for aspiring security professionals
Raspberry Pi Security Projects That Actually Solve Problems
From network monitoring to physical security—practical Raspberry Pi security projects like Pi-hole, VPN gateway, and honeypots without breaking the bank
Continuous Learning in Cybersecurity: Strategies That Work
How I stay current in a field that changes daily—practical strategies including lab exercises, research tracking, and community engagement without burnout
Automating Home Network Security with Python and Open Source Tools
Automation scripts and tools I built to keep my home network secure, including Ansible playbooks, Python monitors, and automated patching systems
I Built an AI Agent to Debug My Homelab: LLM-Powered Incident Response with AIOpsLab
Automated homelab incident response with an LLM agent, reduced diagnosis time from 30 minutes to 5 minutes by auto-correlating Prometheus, Loki, and Tempo telemetry.
Privacy-Preserving AI Training Across My Homelab: Federated Learning with Granular-Ball Computing
Training image classifiers across 4 devices (Dell R940 + 3 Raspberry Pis) without sharing raw data. Granular-ball segmentation reduced network transfer by 82%.