Interesting Projects & Blog

Standalone LLM chat interfaces fall short in enterprise production: they lack deterministic guardrails, risk hallucinations, and cannot reliably update mission-critical databases without supervision.

Real business value comes from uniting hardened software architecture with the cognitive capabilities of AI:

Discuss Automation Opportunities for Your Team

About the Application

As part of my experiments in the App Store section, I built My Learning Style (Můj styl učení). It is an engaging quiz designed for children and students to discover their primary sensory learning preferences – whether they are predominantly Visual, Auditory, or Kinesthetic.

Users answer situational lifestyle questions and instantly receive a visual score chart accompanied by actionable, encouraging study advice.

Technical Architecture: Client-Side Reliability

Rather than introducing fragile external API calls, latency, and client-side credential exposure, the app utilizes an AI-designed evaluation matrix:

• AI-Engineered Matrix: AI was used during the design phase to synthesize pediatric psychology insights into a structured evaluation matrix with child-tailored feedback and practical tips.
• 100% Client-Side Execution: Zero server costs, zero external API keys in client JavaScript, and instantaneous offline-capable evaluation.
• Smooth UX: Includes a subtle 1.5-second processing simulation to give children an authentic feeling of analytical evaluation before presenting the results.
• Lightweight Stack: Modern ES Modules, Tailwind CSS, and Lucide icons.

Project Background: Isometric Liberec (izomapa.fun)

The Isometric Liberec (izomapa.fun) project was created by my former colleague and friend Adam Klatovský. The project takes real open geographic data (from Overture Maps, OpenStreetMap, and digital elevation models from ČÚZK) and algorithmically converts them into a rich isometric pixel-art map reminiscent of classic SimCity 2000. It features full seasonal transitions, dynamic day and night cycles, and smooth deep-zoom navigation.

The technical concept caught my attention, so I joined the project and contributed a series of pull requests. I began advising on architecture to ensure long-term sustainability. A key contribution was helping optimize the map builder and pipeline architecture, allowing multiple enthusiasts to collaborate on spatial datasets concurrently without suffering "merge hell" and making it easy to support each other when integrating new territories.

The complete codebase is open-source on GitHub: github.com/adamklatovsky/izomapa.

---

Expanding to Regional Towns: Dětřichov, Heřmanice and Mníšek-Fojtka

Leveraging the project and the improved builder pipeline, I processed the spatial datasets and generated interactive isometric maps for three regional communities: Dětřichov, Heřmanice, and Mníšek a Fojtka.

All three maps are published on their own subdomains and can be explored in any browser:

Under the Hood: Technical Architecture

• Geodata Processing: Extracting raw building footprints, road vectors, and waterways from Overture Maps and OpenStreetMap, merged with high-resolution digital relief models.
• Isometric Projection: Dimetric/isometric 2:1 angle without vanishing points – lines remain parallel, preserving true scale across the entire canvas.
• DeepZoom Pyramids with WebP: Rendering huge tens-of-thousands-pixel maps sliced into hierarchical WebP tile pyramids for instantaneous browser loading.
• OpenSeadragon Renderer: Interactive multi-layer viewport with smooth seasonal transitions and procedural detailing (procedural tree placement, lighting, and ambient shadows).

Huge thanks to Adam Klatovský for creating such an imaginative project and making it open to the community.