Skip to content

AI Expert · Software Architect · CTO

Mariusz
Tarnaski

I help companies implement AI where it delivers real business impact: in processes, products, and software development teams.

AI implementation for business

I help companies find AI use cases that truly increase revenue, cut costs, and streamline processes. From concept to production deployment.

Developer team training

A systematic approach to working with AI. Your developers accelerate their work and deliver better code from day one.

Building AI products

I integrate AI directly into your products and processes to strengthen your business, improve operations, and increase revenue.

offer

AI for business

I identify where AI will deliver real value for your company and implement solutions: from prototype to production.

  • Discovery and use case selection
  • Proof of Concept and prototypes
  • RAG and enterprise data integration
  • Chatbots and AI assistants
  • Voice assistants
  • Process automation
  • Production deployments

AI for developers

I train development teams to work with AI and implement standards that make AI truly accelerate software development.

  • Practical AI-assisted programming methods
  • Claude Code adoption
  • AI-powered code review
  • AI-assisted project documentation
  • Governance and security
  • AI agent workflow standards
  • AI-first organizational transformation
Learn more

BMad: AI-driven projects

A structured approach to working with AI. Specialized AI agents guide the project through every phase: from planning to implementation and code review.

  • BMad method in practice
  • Specialized AI agents for each phase
  • Jira and GitHub integration
  • Full control over quality and architecture
Explore BMad

case studies

Selected projects where AI is the subject rather than the tool they were built with: an assistant inside the app, a connection over MCP, a machine learning model of its own, plus tooling and rollouts for teams working with AI.

Developer team transformation

  • Training
  • AI Workflows
  • BMad

End-to-end AI adoption for development teams. I define AI agent workflow standards, establish code review practices, prompt patterns and guardrails, select tools, and build daily workflows. Developers in these companies prove every day that AI is capable of delivering ambitious projects.

Headroom AI: Claude Code and Codex limits in the menu bar

  • Open Source
  • macOS
  • Swift
  • DevTools

A free macOS app I wrote for myself and published openly. Claude Code and Codex only ever show the account you happen to be logged into, so with a dozen subscriptions, finding one that still has room means switching and checking over and over. Headroom polls every account at once and shows, in a single panel, how much is left in each limit window and when it comes back.

Helpi: service companies and parts suppliers in one system

  • SaaS
  • Web App
  • Mobile
  • STT
  • AI Assistant

A platform that works on both sides of a service job. Service companies manage clients, facilities, work orders, and recurring maintenance with a full repair history. The technician in the field simply dictates notes into their phone, and AI transcribes the audio and video, then uses it to suggest which parts to order from a supplier. The supplier keeps working in their own email inbox, enhanced with a management layer: assigning emails to people, statuses, and deadlines. AI analyzes each inquiry and proposes ready-to-send reply drafts with prices from the product catalog, and technicians can search that same catalog themselves right from the Helpi app.

amiready.app: financial forecasts for years ahead

  • FinTech
  • Web App
  • MCP Server
  • SaaS

A web tool for planning major life decisions by modeling your finances for years ahead. You enter income and expenses, and the app shows projected cash flow and balances up to 10 years ahead. You compare “what if” scenarios and get alerts about months when things may get tight. You can talk about your finances in natural language in your favorite AI tool thanks to MCP integration.

Product recognition for UAE government

  • Computer Vision
  • OCR
  • Automation

Visual content recognition system for product label photos, built for an international standardization organization on behalf of the UAE government. The system identifies the product, class, manufacturer, composition, and certifications from photos, then automatically fills registration forms. A process that took days was reduced to minutes.

RAG for industrial robotics manufacturer

  • RAG
  • Embeddings
  • Vector DB

A knowledge search engine that processes gigabytes of technical documentation for an industrial robotics manufacturer. The system delivers precise answers to engineers' questions based on internal company databases, drastically reducing information lookup time.

Virtual influencers

  • Persona AI
  • RAG
  • Voice
  • TTS

A platform for creating virtual avatars of well-known personalities. Based on materials from YouTube, TikTok, and other platforms, we build the psychological profile, speaking style, and knowledge base of each person. Users chat with the avatar, which replies like the real influencer, with their knowledge and personality. Every response can also be listened to in a generated voice of that person. An ideal tool for learning from authorities in any field.

Nupi: AI orchestrator

  • Go
  • Rust
  • Open Source
  • CLI

An intelligent assistant that supervises and coordinates multiple AI programming tools simultaneously. Manage Claude Code, Codex, and other tool sessions by voice or from your phone, even when away from your computer. Nupi automatically assigns tasks, monitors progress, and reports results.

  • Open Source
  • ACP
  • E-commerce
  • PHP

The world's first PHP library for OpenAI's Agentic Commerce Protocol. It enables selling products directly in ChatGPT and other AI tools. A new customer acquisition channel that will revolutionize the market.

BrickHunt: LEGO set recognition

  • Computer Vision
  • Mobile
  • ML

The fastest way to find the cheapest deal on a LEGO set. Just point your phone camera at the box on a store shelf, and the app recognizes the set in under 2 seconds and compares online prices. The ML model runs locally on the phone, with no internet connection required.

Other projects, not directly related to AI, but equally ambitious.

Whisma: a desktop app for Gmail

  • Desktop App
  • Gmail
  • SaaS

A desktop app that lets you comfortably work with multiple Gmail accounts in one place. No more juggling dozens of browser tabs, all your accounts are at hand. With the multi-panel view you can easily work with multiple emails at the same time. Instant switching between accounts, system notifications, global keyboard shortcuts, and system tray integration. Available on macOS, Windows, and Linux, in 37 languages.

Centered Volume: a centred volume HUD for macOS

  • Open Source
  • macOS
  • Swift
  • CoreAudio

A free macOS app I wrote for myself and published openly. In macOS 26 Tahoe the system volume overlay moved to the top-right corner, which on a wide or multi-display desk means hunting for it. Centered Volume draws its own panel in the middle of the screen, above the dock, and lets you switch output with one click instead of a trip to Settings.

MyBrickHeadz: a LEGO BrickHeadz fan site

  • LEGO
  • Hobby
  • Collection
  • Fan Site

A site made out of passion, dedicated to the LEGO BrickHeadz series, which I collect and love. I tracked down many details about how the series came to be, the designers' behind the scenes work, and the later stories of individual sets. The site features a catalog of over 200 figures, trivia about rare sets, news, a microblog, and my complete collection.

  • SaaS
  • Docker
  • Tunneling
  • DevTools

A tool for securely sharing local projects over the internet. With a single Docker command, you create a tunnel that lets clients, testers, or other developers see your local service at a public URL. Ideal for demos, mobile testing, webhooks, and team collaboration without deploying to a server.

  • Docker
  • Course
  • DevOps
  • Education

A Docker course designed together with Łukasz Lach, Poland's first Docker Captain. 13 hours of video content, 227 lessons. Modules tailored to specific technologies: PHP, Python, Node.js, MySQL, PostgreSQL, MongoDB, Nginx, Redis, and more. Each lesson includes ready-to-use configurations for your own projects.

media and talks

Conference talks

about me

Technology has fascinated me since childhood. I wrote my first code at age 8. For over 25 years, I’ve been building software, teams, and technology companies. I’ve progressed from developer, through Project Manager, to CTO and co-owner of a software house.

I know what technology looks like from the inside and how it translates to business results. Over the years, I’ve managed development teams, built them from scratch, conducted over 100 technical interviews, negotiated budgets, and designed solution architectures. I can bridge the gap between technology and business to help you make the right AI decisions.

I’m equally comfortable discussing system architecture with developers and deployment strategy with the board. I’ve spoken at multiple conferences and conduct training on AI implementation in companies.

I lead AI initiatives end-to-end. I identify where AI will deliver the greatest value, design solutions from scratch with AI in mind, set technical standards, and deploy them in production. I take full responsibility for business outcomes, not just code.

Today, I’m fully dedicated to AI. I integrate applications with language models, build RAG systems, work with voice assistants and local models. My development workflow is AI-first: I mainly work with Claude Code and Codex, and I implement the BMad method not only in my own projects but also in the teams I train.

Languages and frameworks

  • Python
  • TypeScript
  • Node.js
  • React
  • Go
  • Rust
  • PHP
  • Next.js

AI and ML tools

  • OpenAI
  • Anthropic
  • Claude Code
  • Codex
  • Gemini
  • LangChain
  • CrewAI
  • n8n
  • TensorFlow
  • OpenCV

Databases

  • PostgreSQL
  • MySQL
  • MariaDB
  • MongoDB
  • ElasticSearch
  • Redis

Infrastructure

  • Docker
  • Kubernetes
  • AWS
  • Azure
  • GCP

contact

Looking for support in implementing AI in your company or team? Let’s talk about how I can help.