ICT Infrastructure Engineer · Developer Tools Creator · Open Source
Workplace Engineer @ novosys.ch
Building tools for practical problems. Everything runs local. Everything offline-first.
- Network: Ubiquiti UniFi (Cloud Gateway, UDM Pro, WiFi 6/7)
- Automation: Home Assistant, Raspberry Pi, Zigbee, Matter, Z-Wave
- Storage: Synology/UGREEN NAS, Microsoft 365, Exchange Online
- Devices: Apple Silicon (macOS), Surface Laptop 7 (Snapdragon X Elite, Windows 11), Linux servers
| Layer | Technology |
|---|---|
| Desktop Apps | Rust + Tauri, C# + WPF (.NET 8), Swift (macOS/iOS) |
| Web/Frontend | TypeScript, React, Tailwind CSS |
| Backend | Rust async (Tokio), Python, PowerShell 7 |
| Data | SQLite with FTS5, PostgreSQL |
| AI | Ollama (local), Qwen 3.5, GitHub Copilot, Windows Copilot |
| IDE / Tools | Visual Studio 2022, Visual Studio Code, Xcode, Windows Terminal |
| DevOps | Git/GitHub, Docker, GitHub Actions |
Working daily across the full Microsoft stack on a Surface Laptop 7 with Snapdragon X Elite:
| Area | Tools & Services |
|---|---|
| Hardware | Surface Laptop 7, Snapdragon X Elite, Windows 11 on ARM |
| Development | Visual Studio 2022, Visual Studio Code, Windows Terminal, PowerShell 7 |
| AI Companions | GitHub Copilot (IDE integration), Windows Copilot (native ARM) |
| Enterprise | Microsoft 365, Exchange Online, Entra ID, Microsoft Teams |
| Cloud | Azure Monitor, Microsoft Sentinel, Microsoft Graph API |
| Project | Description |
|---|---|
| LifePlanner | Offline AI life planner: events, tasks, goals |
| LifeSort | AI-powered file organizer: rules, duplicates, automation |
| CleanFlow | Smart disk cleanup: temp files, caches, duplicates |
| ClarityDesk | Universal display interpreter: OCR + local AI |
| DeviceHealth | Cross-platform system health monitor |
| MailPilot | Offline-first email client with AI classification |
| BugRadar | Real-time log analysis & AI-driven incident detection |
| Project | Description |
|---|---|
| LogLens | AI log aggregator: clustering, anomaly detection, FTS |
| StateForge | State machine analyzer: parse, visualize, generate |
| agent-governance-console | AI agent governance, tracing & audit: Axum, OpenTelemetry, Azure Monitor / Sentinel |
| private-model-orchestrator | Privacy-first foundation model orchestration for Apple device fleets |
| Project | Description |
|---|---|
| SwiftAgent | Native iOS AI chat interface: local model integration |
| CodeWhisper | macOS AI code assistant with NSServices integration |
| Project | Description |
|---|---|
| NetSweep | Network storage audit & cleanup: NAS, SharePoint, DFS, Windows |
| NetDashboard | Network & mail diagnostics: DNS, M365/Exchange Online, Windows |
| Project | Description |
|---|---|
| SiliconMark | Apple Silicon LLM benchmark suite: Token/s, RAM, ANE |
| NetScanX | Cross-platform network discovery & diagnostics CLI |
| GardenFlow | Modular home garden automation (MQTT, FastAPI, Docker) |
| HomePortal | Self-hosted home services portal (FastAPI, Docker) |
| azure-cost-forecasting-engine | Azure cost analysis, forecasting and optimization recommendations via Consumption API |
| eventhub-otlp-mapper | Map Azure EventHub messages to OpenTelemetry Traces and Metrics (OTLP, JSON/Avro/Protobuf, Azure Monitor) |
| Project | Description |
|---|---|
| M365-AdminTool | PowerShell toolkit for Microsoft 365 administration |
| entra-access-graph-engine | Entra ID privilege access graph: detect escalation paths, hidden admin chains, and risk scores (Rust, Graph API) |
| entra-least-privilege-analyzer | Read-only Rust CLI for Entra ID privilege analysis, PIM gap detection and role overlap reporting |
| azure-policy-drift-detector | Read-only Rust CLI to detect Azure Policy drift, prioritize non-compliant resources and generate remediation reports |
| github-actions-security-sandbox | Static analysis and attack simulation for GitHub Actions workflows: injection, Pwn Request, unpinned actions |
Running a local AI lab on Apple Silicon (laptop and desktop), testing open-source models and comparing them against cloud APIs, integrated directly into my own tools.
Local inference via Ollama:
- Daily driver: Qwen 3.5 as the primary model and personal benchmark baseline
- Continuously testing new releases against it: LFM 2.5, Gemma 4, Qwen 2.5 and others
- SiliconMark (my own tool) to measure Token/s, RAM usage and ANE activity per model and runtime
Autonomous agents:
- Nous Hermes running as a 24/7 local agent for long-term workflow and reliability tests
- Handles multi-step pipelines, tool calling and unattended tasks entirely on-device
- qwen3-coder tested as a local coding agent against cloud-based coding tools
- Claude API used as the cloud reference point in agent and workflow benchmarks
- GitHub Copilot integrated in Visual Studio and VS Code for enterprise development: .NET, WPF, PowerShell scripting, M365 automation and code reviews
- Windows Copilot on Surface Laptop 7 (Snapdragon X Elite, ARM-native): daily companion for drafting, research and hands-on testing of Microsoft AI features on Windows on ARM
Integration & testing:
- All my tools with AI features (LifePlanner, LogLens, MailPilot, CodeWhisper and others) support both local Ollama models and cloud APIs
- The goal: find where a well-tuned local model is good enough and where cloud capability actually justifies the tradeoff
- Rust: Tauri desktop apps, async backends, CLI tools
- Python: automation, benchmarks, data processing
- C#: Windows desktop and enterprise tools with Visual Studio 2022
- Local AI: Ollama, Qwen 3.5, offline-first
- macOS/iOS: Swift development, Xcode
- Microsoft / Azure: M365, Exchange Online, Entra ID, Azure Monitor, WPF, Visual Studio and enterprise integrations
- Windows on ARM: Surface Laptop 7 with Snapdragon X Elite as the primary Windows development machine
20+ years in logistics & supply chain management
3+ years in ICT infrastructure & support
Certifications: SIZ Informatik, ICT Power-User, Leadership
Translated from logistics mindset: systems thinking, process optimization, hands-on engineering.
- Website: raystudio.ch
- Work: novosys.ch
All repos are open source (MIT license). Issues & PRs welcome.