Danish developer (born 1975) building practical, end-to-end systems with full-stack ownership.
- System-builder mindset
- Focus on working solutions over theory
- Lightweight, cost-efficient architecture
- End-to-end delivery: frontend + backend + infra
- Values clarity, speed, and ownership
- Languages: Python (primary), JavaScript, HTML/CSS, COBOL, C++, Pascal, Java
- Backend: Flask, SQLite, REST APIs, MQTT, Home Assistant integrations
- Infra/DevOps: Docker, Google Cloud Run, AWS (EC2/Lightsail/Route53/Terraform experiments), GitHub Pages
- Environment: Windows 11, VirtualBox (Home Assistant VM), DuckDNS, networking/port forwarding
- AI/ML: PyTorch (GPU), DQN, local LLM workflows (llama-server/Ollama), RSS AI summarization and scoring
- RTX 3070 Ti (CUDA 12.x)
- Zigbee2MQTT + Sonoff Zigbee dongle
- IoT sensors (energy, temperature, VOC)
- Solar panel + battery metrics
- Garmin watch integration experiments
- AI NewsReader (
othar.dk) AI-powered RSS aggregation, translation, opinion-vs-fact separation, prioritization logic, optimized for speed and low cloud cost.
Live app:newsreader.othar.dk(Google AI Studio app). - Kenobi (Dog Walk Tracker)
Home Assistant + MQTT +
input_datetime, shareable web app (othar.dk/kenobi), intentionally minimal data model. - iPad Kitchen Dashboard Lightweight, performance-aware dashboard optimized for older iPad hardware.
- DQN Car Simulation Deep Q-learning with replay memory, trajectory optimization, GPU training, later ported to Java for speed experiments.
- Global Meta-DB Concept JSON-based lightweight shared data layer across micro apps with expiry metadata.
- Home Assistant (VM)
- Zigbee2MQTT
- MQTT broker
- Dockerized services
- Energy data ingestion pipelines
- Nilan gateway experiments
- Eufy security integrations
- Custom health-check endpoints
- Log-driven debugging
- Reproducible, copy/paste-ready workflows
- Speed + simplicity first
- Deliberate daemon/web separation
- Cost-aware cloud decisions
- Iterative building with practical AI integration
- Local models when possible (reduced SaaS dependency)
- Domain:
othar.dk - GitHub: @RasmusOtharKirketerp
- Public demos, AI media experiments, and structured agent prompting
Legacy enterprise background (COBOL/DB2 mindset) combined with modern Python/AI, IoT automation, GPU experimentation, and lightweight cloud infrastructure for real daily-use tools.
Golden Retriever owner (Kenobi), long-term builder mindset, and interested in truth-focused AI prompting.