Focus
AI-assisted software development – faster, without giving up control
Daniel has used AI tools in development since 2022 – first in teaching, since February 2026 with coding agents in the product. His principle: the AI helps write, tests and human review decide.
What matters to him is not the tool but the process around it: tasks as issues, traceable reasoning, automated checks and a review that is taken seriously. That way a team gets faster without quality and security suffering.
He takes teams through this shift – with clear rules rather than enthusiasm alone.
Evidence from his career
uCORE Systems
Proposed Claude Code in a retrospective and switched the process within two weeks: GitHub issues as the backlog, agents in containers, tests and review as the gate. Architecture and review stayed with him, the agents did the implementation.
Coding bootcamp for career changers
AI in teaching since 2022: teaching material on AI-assisted software development (documented in the employment reference), curriculum revised and standardised with ChatGPT.
What references say
“Integration of artificial intelligence into the production of teaching material, and creation of teaching material on AI-assisted software development.”
“He teaches cutting-edge software development in Java, proven patterns and processes, bundled with the use of the latest AI tools.”
“His curiosity for new technologies, frameworks and programming approaches is remarkable.”
Related projects
Clauboy
Built in March 2026 to run Claude Code agents per GitHub issue in isolated Docker containers with their own git worktree – state via labels, prompt-injection mitigation, tokens in the OS keychain. About 400 issues worked through it in production. Since superseded by newer tools; remains evidence of early, independent agent orchestration with a security mindset.
This website
Interactive CV as a RAG application (retrieval-augmented generation): the assistant answers questions from the page content – profile, references, prepared building blocks –, highlights matching stations and reveals details that are hidden at first. Most questions are answered without calling the model. No framework, under 100 KB, load time in milliseconds. Built with Claude Code; the source code stays private for now and is shown on request.
Edge controller on Raspberry Pi CM5
Status display via WS2812B LED strips over SPI and speech output via ALSA on the edge controller of the care-IoT product. Plus ESP32, ATmega and ATtiny experience – oscilloscope and 3D printer included.
Technologies
Explore in the conversationTo the interactive CVRequest