AI developer based in Wrocław, Poland
I bring AI into businesses. I build assistants that answer calls and book appointments, workflows that sort enquiries and documents on their own, and applications built on language models. I handle all of it, from the first conversation through launch to ongoing care.
An animated console shows how I work with AI agents. Claude Code adds a feature to the Fonvel assistant, tests and deploys it, and checks the status of my services. Codex builds a workflow that classifies enquiries with an LLM and saves them in the CRM, then adds a price list to the assistant’s knowledge base and checks its answers. After the animation you can type commands, for example help.
A demo scenario. Values marked “live” come from my servers.
I connect language models to the tools you already use: email, calendar, CRM and documents. The workflow does the repetitive work, and people approve the decisions that matter.
Example flows with made-up data. This is what the solutions I build look like.
Handling enquiries, documents and reports where a language model does the repetitive work and a person approves the decisions. Connected to your email, CRM, calendar and spreadsheets.
OpenAI and Anthropic APIs, tool calling, structured outputs, Python, webhooks
Assistants that talk to customers on the phone and on the website, answer from the company’s knowledge and book appointments in the calendar.
LiveKit, RAG with pgvector, speech recognition and synthesis, calendar integrations
See it in practice: Fonvel, ChatbotAssistant
We start by finding where the most time goes. I pick one process, build a prototype on your data and then roll it out with control over cost, answer quality and data security.
process review, prototype, evals, cost monitoring
From an idea to a working application: backend, frontend, sign-in, payments, deployment and operations. That is how all my products were built.
Python, FastAPI, Go, Rust, Next.js, React, PostgreSQL, Stripe, Docker
See it in practice: PlanMyTeam, Fonvel
Each one is live, so you can open it and try it. I am responsible for all of it: the product, the code, payments, deployment and operations.
An AI phone assistant for businesses.
It answers calls when nobody at the company can pick up. It replies from the company’s offer and knowledge base, books appointments in the connected calendar and leaves a short note after every call. Each business gets its own workspace and a versioned assistant configuration.
Built with: Next.js, Go, PostgreSQL + pgvector, LiveKit, Stripe
AI training plans for football coaches.
A coach describes the session in a chat: age group, topic, number of players and available equipment. PlanMyTeam picks drills that follow the Polish FA (PZPN) curriculum from U7 to seniors, draws animated pitch diagrams and lays out a ready two-page A4 session plan.
Built with: Rust, React, TypeScript, LLM, PDF
A voicebot, chatbot and booking system in one.
It answers customer questions on the website and over the phone, works out what they need and books appointments according to the company’s schedule. Available in 20 language versions.
Built with: FastAPI, PostgreSQL, Redis, Authelia, Voice AI
A smart home that runs on a single ESP32-S3 microcontroller.
Automations, the control panel and the ESP-NOW mesh network run locally, with no cloud, no external server and no subscription. MicroAssistant is sold as a one-time licence or as a ready, pre-programmed device.
Built with: ESP32-S3, ESP-NOW mesh, Zigbee, MQTT, FastAPI
One desk, every device.
A software KVM for Windows and Android that works only on the local network. Cursor, stylus, clipboard, files and screen view are separate permissions, each granted deliberately.
Built with: Windows, Android, mTLS, mDNS
An encrypted peer-to-peer messenger.
Text channels, voice and video calls and file transfer with end-to-end encryption (MLS). Whenever possible, devices connect directly with no server in between. Runs in the browser, on Android and on desktop.
Built with: Rust, WebAssembly, WebRTC, MLS
My products and this site. They are checked automatically every minute, and the line next to each service shows the last hour.
Response radar: the closer to the centre, the faster the service responds.
The technologies I use, linked to the projects they actually run in. Hover a point to see its links and drag to rotate the map.
Libraries and tools under the MIT and Apache 2.0 licences. DevBridge makes working with AI agents easier; the rest are building blocks for MicroPython on the ESP32 that I write after hours.
I’m Mateusz Sury and I live in Wrocław, Poland. My main focus is practical AI: I bring AI into businesses, build workflows on top of language models and create assistants that talk to customers.
I don’t stop at a prototype. I design the architecture, write the code, connect integrations and payments, deploy and keep everything running. I work with AI agents such as Claude Code and Codex every day. They help me deliver faster, but I own the quality myself: tests, code review and monitoring after launch.
After hours I write open-source libraries for the ESP32 and MicroPython, which is where the IoT in my portfolio comes from.
Describe the process or idea you want to improve with AI. I’ll reply to the email address you enter in the form.
Wrocław, Poland