Flutter
Unity
Infoshare
FlutterE2E
Aplikacja konferencyjna dla eventu IT — agenda, mówcy, networking i obsługa offline.
AI-Augmented Mobile Developer4+ years of commercial development. I work with AI agents every day.
I ship software faster because I work with AI every day — not just on holidays.
I ship commercial code in Flutter and Unity. Over the last year I rebuilt my workflow around AI agents — Claude Code with my own configuration, hooks and skills sits in my projects every day. I'm looking for a role or contract where this acceleration matters.
- Production code in Flutter (Dart) and Unity (C#)
- Claude Code: my own config, hooks, skills, subagents
- BSc in Computer Science · Technical English B2/C1
How I work — AI in my daily workflow
Four concrete areas where AI sits in my daily toolkit. It's not about pasting questions into a chat — it's a full workshop: agents in a loop, custom tools, hooks that enforce project standards, automations around review and CI.
AI-assisted development as a daily practice
- Claude Code as a daily driver — I use it for most programming tasks: implementation, debugging, exploring unfamiliar codebases, refactoring.
- My own configuration — settings, permissions, custom slash commands, skills and subagents tailored to my projects.
- Working with the agent in a loop — planning the task, delegating exploration to subagents, verifying output, iterating. Not "chat" — a controlled workflow.
- Awareness of model limitations — I know when an LLM is wrong, how to detect it, and how to set up the process so errors don't leak into production.
Agents and scripts supporting the SDLC
- Specialized subagents — code reviewer, code explorer, planner, tester. Each with a different system prompt and a different tool scope.
- Project hooks — enforcing conventions, blocking risky operations, automatic formatting/linting before commits.
- MCP (Model Context Protocol) — using off-the-shelf MCP servers to integrate with team tools.
- Slash commands and skills — they capture repeatable team workflows (PR review, generating tests, preparing release notes).
Automations around build and CI/CD
- Triage of errors and regressions — the agent reads CI logs, suggests the cause, sometimes proposes a fix.
- Automatic change descriptions — generating readable release notes / changelogs from diffs and commits.
- AI-assisted PR reviews — a first layer of checks (style, obvious bugs, missing tests) before a human sees it.
- Automating repetitive repo operations — bumping dependencies, generating scaffolds, codemods, large refactors with verification.
Generating code and end-to-end features
- "Brief → working module" pipeline — with a clear verification step (tests, type checks, manual review) before merge.
- Component / widget generators — following project conventions, not in a "whatever" style.
- Codemods and migrations — large-scale code changes driven by the agent, with tests pinning behavior.
- Discipline — the agent doesn't merge on its own; every output goes through review and the build pipeline. AI accelerates, it doesn't replace responsibility.
What I can deliver
First, what I have documented experience in. AI tooling and business processes — below, as areas I'm actively expanding into.
Mobile development — Flutter / Dart
- Android + iOS mobile apps from scratch and growing existing ones,
- responsive and scalable UI, custom widgets, asset optimization,
- state management with Riverpod, Freezed, Dependency Injection, Clean Architecture,
- REST API and Firebase integrations, auth, sessions, offline-first data,
- debugging on physical devices, publishing to Google Play / App Store.
- Flutter
- Dart
- Riverpod
- Freezed
- REST API
- Firebase
- Clean Architecture
- Android
- iOS
- Google Play
- App Store
Interactive apps and games — Unity / C#
- games and interactive tools, UI Toolkit / Canvas / responsive UI,
- Firebase, REST API, Unity IAP integrations,
- design patterns: State Machine, Pooling, Scriptable Object,
- separating MonoBehaviour from pure C# logic, attention to readability and standards,
- building editor tools and designer-facing tooling.
- Unity
- C#
- UI Toolkit
- Firebase
- Unity IAP
- State Machine
- Pooling
- Scriptable Object
Common to both stacks
- code readability, architectural scalability, long-term maintainability,
- the full product lifecycle: planning → implementation → testing → release → maintenance,
- collaboration with developers, testers, project managers, clients.
Portfolio
Three competence tracks: commercial mobile apps built at Honeti, my own agentic-programming projects, and an AI workshop — the methodology I bring to every team.
Commercial portfolio —
As a Mobile Developer at Honeti — full lifecycle from planning through implementation, testing, and release to long-term maintenance.
- Flutter
- Dart
- Riverpod
- REST API
- Firebase
- Clean Architecture
- CI/CD
- Unity
Uprawnienia Budowlane
FlutterE2E
Aplikacja do nauki dla osób przygotowujących się do egzaminu na uprawnienia budowlane — testy, materiały, model subskrypcyjny dostępu do treści.
Gastro Ninja Klient
Unitymaintenance
Aplikacja klienta końcowego w ekosystemie Gastro Ninja — platforma do zamawiania jedzenia na wynos i dostawę, w modelu podobnym do pyszne.pl / pizzaportal.
Agentic programming — first-party deployments
AI is part of my production process. Three live projects where you can see it in practice.
Personal mobile app
- Flutter
- Firebase
- CI/CD
In the final stage of production. Full CI/CD pipeline with AI agents in the loop.
Personal OSS — localization tool
- TypeScript
- AGPL-3.0
- Web
Open-source, preparing a public release. Format-faithful architecture, AI translation via the user's own API keys.
buzzards-soft.com
- Next.js
- React
- i18n
The site you are reading, built with the active assistance of AI agents. The entire process is publicly documented.
AI workshop — methodology
What I bring to every team — independent of the project.
Sprint-driven workflow with typed sessions
Every change starts from durable context — BRIEF.md, a sprint file, or an ADR — not from a fresh prompt. Every agent session has a type (planner / implementer / tester / reviewer) and a single declared deliverable. Reviewer sessions are read-only — they cannot edit code, only write a list of concerns. The result: faster, but still disciplined — the agent does not merge on its own.
- BRIEF.md
- CLAUDE.md
- SPRINT_RULES.md
Claude Code configuration
Per-project settings.json with permissions, custom keybindings, hookify rules with a concrete hook, custom status line / powerline.
.claude/settings.json
Specialized sub-agents
code-reviewer, sprint-implementer, sprint-tester, sprint-reviewer, codex-rescue. Each sub-agent has its own tools, prompt, and scope — not one universal agent that does everything.
sprint-implementer / -tester / -reviewer
Custom slash commands and skills
Repeatable SDLC processes packed into commands the whole team can run — from sprint planning to PR review.
/sprint-run · /sprint-plan
MCP integrations in workflow
context7 (live library docs), Playwright (browser UI verification), Maestro (mobile e2e). The agent has real tools in my stack, not just knowledge.
context7 · playwright · maestro
Persistent memory + bilingual rule
Memory system with types (user / feedback / project / reference) + a PL-conversation / EN-markdown meta-convention enforced by the workflow itself. Context does not vanish between sessions.
PL conversation · EN markdown
Full control over my git workflow
The agent works on a worktree or a feature branch, never directly on main. It only opens a PR when I ask. The split is planned: what the agent does, what I do, how it ties into CI/CD. Full control over what lands in git.
worktree · feature branch · PR on demand
Growth
A short list of topics I'm actively going deeper on — mostly to extend my agent toolkit beyond pure production code. I'm signaling direction, not selling it as a service.
- Agents specialized for specific SDLC processes — automatic triage, generating regression tests, codemods at a larger scale.
- RAG pipelines over closed knowledge bases — mainly aimed at internal project documentation.
- Business process automation with an LLM layer — n8n, Make.com — I know the tools by name and am just starting to explore them.
I test some of this on my own projects. If one of these areas is on your team's roadmap — you've come to the right place.
About me
A developer with 4+ years of commercial experience (Flutter, Unity) for whom AI is a workshop, not a curiosity. Focused on shipping software — faster and cleaner thanks to tools I configure myself. Previously 14 years in technical installations — useful for projects that bridge software and hardware.
Most of my commercial work has been Flutter mobile apps (Honeti) — from the first line of code, through REST API and Firebase integrations, auth, data sync and offline work, all the way to store releases and maintenance. I've also worked in Unity on games and interactive apps. In both stacks I put strong emphasis on code readability, architectural scalability, and making sure the project still lends itself to development years in.
Over the last year my workflow has gone through a noticeable shift. I started treating AI agents as a workshop tool, not a curiosity. Claude Code became my daily driver — with configuration tailored to my projects. Things I used to do by hand (review, exploring unfamiliar codebases, codemods, scaffold) now happen faster, but still under my control — the agent doesn't merge on its own, every output goes through build and review.
I don't call myself an "AI Specialist".
I call myself a developer who deliberately uses AI in daily work — and who, thanks to that, ships more.
That's the distinction I care about: my track record is in writing production code — AI is the way that code is now produced faster.
Before I started programming professionally, I worked for 14 years as an electrician and installer of CCTV, alarm, fire safety and smart home systems. For most software projects this background is neutral. On projects that bridge software and the physical world — IoT, smart buildings, automation — it gives me a perspective a pure developer doesn't have.
I value end-to-end work: from analyzing the problem, through implementation, to maintenance. I don't leave projects in a "PoC" state. I don't promise things I can't deliver.
Education and certificates
Wyższa Szkoła Biznesu — National Louis University, Nowy Sącz
- Degree:
- BSc in Engineering
- Major:
- Computer Science
- Specialization:
- Business Application Programming
Zespół Szkół Elektryczno-Elektronicznych, Szczecin
- Degree:
- Technician
- Major:
- Electrician
- Specialization:
- Power Electronics
Additional qualifications
- SEP E + D
- Driving license cat. B
- Sailing license
Open to contracts, full-time roles, and interesting projects.
I'm the best fit where you're looking for a developer who already has working products under his belt and can work with AI in the workflow. If this profile fits what you're looking for — get in touch and let's talk through the details.