5+projects running in production
10+essays on AI engineering
24/7agents and pipelines running

What I build

AI systems that finish the job

Not demos — operations. Every system ships with measurable output, tests, and clear ownership.

01

MCP servers

Tools that other people's agents use. I design and run MCP servers for real workflows — from internal data to automation.

02

Agent pipelines

From GitHub issue to merged PR with no human in the loop. Plan, patch, tests, review — agents drive the whole cycle.

03

Autonomous loops

Long-running workflows that watch themselves: research, monitoring, recurring jobs with checkpoints.

04

Evals & reliability

Anyone can demo an agent. I measure task success, human-intervention rate, and regressions — then tune the system on the numbers.

Selected work

Projects with context and impact

Fewer projects, more depth. Each one runs in production with real users.

ProdejUp — screenshot

ProdejUp

AI chatbot for Czech e-shops — installed in 30 minutes.

SaaS with a custom AI chatbot: Next.js, FastAPI, Stripe, and OpenRouter. Live product with paying customers.

Next.jsFastAPIStripeAI
BazényDnes — screenshot

BazényDnes

Pool-equipment e-shop I co-founded.

Complete e-commerce build from design to operations — catalog, orders, SEO, and content management.

E-commerceSEONext.js
Aster · internal

Aster

Local voice AI assistant — security-first, no cloud.

A personal "Jarvis": local voice control, a supervised browser, and guarded integrations. Privacy first.

Local AIVoiceSecurity
ShipGuard UE · internal

ShipGuard UE

Unreal Engine 5.8 plugin — packaging & hitch doctor.

A tool for game developers: build diagnostics and performance-hitch hunting right in the editor.

Unreal EngineC++Tooling
Cognera ERP · internal

Cognera ERP

Internal ERP system — full-stack work on a production team.

Java, Spring Boot, PostgreSQL, and React. Safe migrations, tests, and long-term maintenance of a real system.

JavaSpring BootPostgreSQLReact

Signature

From request to shipped change. Autonomously.

Agents plan, edit code, and test. The process that holds it together is my know-how.

  1. 1

    Request

    A task or bug report arrives. The system handles it end to end.

  2. 2

    Plan

    The agent gets precise project context and drafts a step-by-step approach.

  3. 3

    Code change

    Several agents work at once — no conflicts, no waiting.

  4. 4

    Tests

    Every change passes a check stricter than a typical review.

  5. 5

    Shipped

    You get a verified result, ready to deploy.

  1. $new request — "add CSV export"
  2. agent: plan drafted (3 steps)
  3. agent: 4 files changed
  4. tests: 128 passed
  5. change approved → deployed

The sequence mirrors a real pipeline run; numbers are illustrative.

My passions

Classical violin
01

Classical violin

Years of training. Chasing precision, emotion, and performance.

Strength / Gym
02

Strength / Gym

Pushing limits. The same stamina needed in code and training.

Video editing
03

Video editing

Crafting stories frame by frame where precision meets rhythm.

Taking projects · Q3–Q4 2026

Have a project? Let’s talk it through.

Send me a few sentences about what you are building. I reply within 24 hours with a concrete next step.

reply within 24 hours