Domain-specific chatbots and web apps with AI at their core. Built on modern LLM architectures and an agentic approach. For small and mid-sized companies, at a price that pleasantly surprises.
My clients usually aren't AI rookies and yet they still entrust their projects to me. ChatGPT, Lovable, Claude, first automations in n8n. It works for a while, then it starts to creak — inconsistent outputs, can't quite ship it, something's always missing or could be better. Sound familiar?
EN / SK / CZSometimes an AI layer on top of an existing system is enough — endpoint, integration, module. Other times it makes sense to build the whole app from scratch with AI at the core. And because AI also speeds up the build itself, you get it for far less than a traditional custom build. Frontend, backend, AI layer, deployment.
AI chatbots that truly understand your business
A domain-specific agent that works with your data, knows your business and decides in context. Built on agentic architecture, modern LLM techniques and a retrieval pipeline tailored to your documents.
AI Agents as the cornerstone of process automation
A workflow built on top of Claude that processes emails, documents, calls and data, and makes decisions in the context of your business. No rigid if-then rules. Real LLM reasoning, wired directly into your tools.
From process to product
Where you live in spreadsheets, tabs and manual steps today, I'll build a web app with AI at the core. It works with your data, makes decisions, learns your business. Frontend, backend, AI layer, deployment — all in one piece.
I work with clients from their very first AI project — and they keep coming back with new ideas. Their loyalty is the best calling card. Ask them yourself!
I came to Jaro with an idea. I wanted an avatar from the future that would help recruit people to the team straight on the website. Today we use AI together to build social content, we create reels with the avatar, and we've sped up our whole communication process.
We had loads of data, but it was scattered across Excel files and systems that didn't talk to each other. Jaro helped us bring it together. He processed the inconsistent data, turned it into interactive dashboards and converted static tables into tools the whole team and our clients use daily.
Jaro joined the project at the moment we needed help and direction. He saved what could be saved and pushed it to a working result. Today we work with him on more AI solutions for our clients.
Jaro is the Swiss army knife of any team. Versatile, sharp and always ready to do something. At Habitable he tackled every challenge with ease and tenacity. Jaro doesn't wait for solutions — he ships them.
Not every problem needs an agentic system with a vector database. Sometimes a well-placed prompt inside an existing workflow is enough. I'll explain to you what we're building and why, so you can live with the solution long-term — without needing to keep me on retainer.
Every project starts with data. Spreadsheets, PDFs, emails, call transcripts, CRM exports, knowledge bases. I take messy inputs and build a structure that LLMs and your team can actually use. Without a solid data layer, any model is just an expensive toy.
A prompt isn't a sentence in ChatGPT. It's a system: a rulebook, context, few-shot examples, validation layers, fallbacks, crisis protocols. An architecture that holds up even when a user types something you didn't expect. Plus evaluation that measures improvements, not feelings.
Processes built as an agent, not as rigid automation. Claude decides on its own when to hit the CRM, when to call the billing API, when to pull a document through an MCP server. It handles emails, calls and documents and reasons in context. Without the limits of no-code.
The questions I get most often before the first call.
I built Shizune as my AI twin. She knows my services, my projects and my working style. Ask her the things you'd ask me — without waiting for a slot in my calendar.
