Senior developer and technical lead building AI tools, developer infrastructure and production-ready web platforms—from architecture and security to deployment.
Not a timeline of commits. A curated set of systems, tools and experiments where architecture, usefulness and production thinking matter.
Architecture discipline for coding agents
A Claude Code plugin that loads an architectural decision ladder into every session and adds focused planning and review commands.
Keeps fast AI-generated code from drifting into needless abstraction, weak boundaries and backwards dependencies.
Concept, architecture, implementation and validation
Trust evaluation for AI skills
A deterministic and LLM-assisted evaluator that grades AI skills across security, quality and hygiene—with evidence for every finding.
Makes third-party agent skills inspectable before they receive access to a developer environment.
System design, trust model, scoring pipeline and implementation
Local AI code review for Drupal
An open CLI review engine combining deterministic security checks with a Drupal-tuned local model and a multi-pass verification pipeline.
Catches Drupal-specific mistakes that generic AI reviewers frequently miss, while keeping source code local.
Product direction, architecture, fine-tuned review workflow and CLI
Local command center for Claude Code
A local-only workspace for finding, reading and running Claude Code sessions across projects from one secured Chromium window.
Turns scattered JSONL histories and terminal sessions into a navigable, persistent engineering workspace.
Architecture, security model, product UX and implementation
Developer-machine attack-surface scanner
A read-only audit of credentials, hooks, MCP servers, agent skills, IDE extensions and supply-chain indicators reachable by coding agents.
Answers what an AI coding agent can access on a machine—and how damaging a compromise could be.
Threat model, audit scope, checks and reporting
Conversational learning from video
An AI learning system for chatting with individual YouTube videos and curated channels using retrieval and real-time responses.
Makes long-form video knowledge searchable, discussable and reusable instead of passively consumed.
Product concept, end-to-end architecture and implementation
Showing 6 systems selected for depth, not recency.
This is not blind code generation. AI accelerates the work; I remain responsible for requirements, boundaries, trade-offs, security and the behavior of the finished system.
Trace the real problem, constraints and data flow before choosing a solution.
Define boundaries, interfaces, risks and the smallest architecture that can hold.
Turn the design into narrow tasks with observable acceptance criteria.
Use coding agents to accelerate implementation without outsourcing decisions.
Review behavior and code, test failure paths and check security assumptions.
Deploy, observe, maintain and simplify the system after it meets reality.
A broad stack is useful only when it supports clear decisions across the full path from idea to operation.
From the shape of the data to the shape of the deployment.
Agents as an acceleration layer inside a controlled workflow.
Deep backend experience connected to pragmatic modern interfaces.
The path from a working repository to a system someone can depend on.
I care about the whole path: from understanding a problem and designing the system to shipping, observing and maintaining it.
I am a senior developer and technical lead from Poland. I have worked professionally with Drupal since 2014, with a background rooted in PHP, Linux, Docker and web architecture.
Today I also build AI tools, developer infrastructure and small software products independently—often in short, focused sessions. That constraint has made clarity, good decomposition and useful automation central to how I work.
I am most useful where product thinking, architecture and practical implementation need to meet.