For eighteen years I built the hosting company Hosting90, up to its sale to WY Group. As CTO I designed the architecture of three production AI products, and I contribute to the open-source inference engine vllm-mlx. Today I build my own products — self-healing data feeds, and Llamat, which measures how AI talks about brands.
Open to selected projects and a long-term role
Prague / remote · Czech, English
Structured product data from any website — prices, specs, EANs — with guaranteed accuracy and zero maintenance on your side.
Fast local parsers, two independent sources with cross-validation, AI checks on every extraction, and a 0–1 confidence score. When a site changes its layout, the parser repairs itself. It can also match the same product across shops.
No dashboards — clean structured data delivered to S3, SFTP, a webhook, or a REST API.
A free AI visibility checker: see how ChatGPT and Gemini talk about your brand and which competitors they recommend instead.
The report shows whether AI mentions and recommends you, where you're missing, who gets named instead, and what to improve. Built for SEO and GEO specialists, agencies, and marketing teams.
Report delivered by email.
An open-source inference engine for large language models on Apple Silicon — 1,600+ GitHub stars.
58 merged PRs: KV cache quantization, constrained decoding, and the Anthropic Messages API compatibility layer that makes vllm-mlx work with Claude Code and OpenCode.
Python · MLX · Apple Silicon
A checkout camera that lets clearly adult customers through without ID and asks everyone else for it — all computed on the device.
Face detection, age estimation, and ID document OCR in real time on a Raspberry Pi 5 with a Hailo-10H accelerator: 11–14 fps versus 1.6 on the CPU alone. I fine-tuned my own age model on manually verified data and quantized it for Hailo — about 12 ms per face, with 99.2% decision agreement with the full-precision model.
Python · PyTorch · ONNX · Hailo · OpenCV · FastAPI
Three production AI products where I built and led the technical architecture and AI stack. Past role — not products of mine today.
AI-powered competitive intelligence for marketing agencies.
Agents auto-tag ads, extract hooks, classify CTAs, and tag creatives — all as reliable structured outputs.
Stack: Qwen 3.6 on vllm-mlx (Apple Silicon M3 Ultra). A batch-friendly workload with reliable structured outputs — owned inference made sense economically and operationally.
Led the AI stack and agentic flow of the project as CTO of production AI products.
Shoptet e-commerce analytics for online merchants.
AI agents identify margin leaks, recommend pricing changes, auto-tag transactions, and run multi-step orchestration over orders, shipping, ad costs, and returns.
Stack: Google AI (Gemini). Chosen deliberately — Margly needed low-latency responses for multi-step tool calling and autonomous orchestration. Running inference onsite wouldn't have been economically viable at this task class.
Led the AI stack and agentic flow of the project as CTO of production AI products.
Customer intelligence — mines Reddit, Hacker News, and Product Hunt for pain points, trends, and market gaps.
Discovery and classification agents surface signals at high volume; summarization agents distill the nuance worth acting on.
Stack: hybrid orchestration. Discovery and classification agents on Qwen 3.6 / vllm-mlx (high-volume, batch-tolerant); final summarization and nuance-heavy reasoning on Google AI where the per-token premium was justified by output quality. Routing decided per agent task.
Led the AI stack and agentic flow of the project as CTO of production AI products.
I founded Hosting90 in 2002. For eighteen years I built it from a garage through a 25-person team to a successful international exit to WY Group in 2020. I know cash flow, margins, running a company, and how hard it is to earn every koruna. I look at technology not as a programmer but as an entrepreneur.
At the same time I go to the absolute depths of the technology. I'm a core contributor to the open-source project vllm-mlx (AI inference on Apple Silicon), with 58 merged PRs. As CTO I designed and led the technical architecture and AI stack of three production AI products (Advanty, Margly, Discury).
These products show how technical decisions follow the economics: the batch-friendly work in Advanty ran entirely on owned infrastructure (Apple Silicon / vllm-mlx) at a fraction of the cost, Margly on the cloud for orchestration reliability, and Discury combined both. When needed, I go down to the network layer — for web data collection I run my own infrastructure built on LTE modems.
I work best on systems where reliability and running costs decide: LLM inference, data pipelines, integrations, and agentic systems. I write the code myself and take things all the way to production. Alongside my own products, I take on selected projects.
AI chatbot for automating customer support in e-commerce. Resolves up to 98% of inquiries without a human, recommends products, and actively helps close sales. I designed and fully led the technical architecture, including integrations into Shopify, WooCommerce, Magento, PrestaShop, and OpenCart. As co-founder I continue to actively contribute to its technical development and operations.
→ lobot.chatComprehensive B2B monitoring dashboard for manufacturers and distributors. GuruWatch tracks partner stock levels and pricing across e-shops in real time, watches price trends, and sends immediate alerts. Clients include brands like Lenovo, Niceboy, and Infinix. I designed and fully built the entire data and scraping pipeline that runs reliably at scale. Day-to-day operations have been handed over to a new owner, but I continue to provide strategic technical support and consulting for GuruWatch.
→ www.guruwatch.cz2H Service (Vapestyle)
A company with 100M+ CZK revenue, 10+ e-shops and brick-and-mortar stores. Under my mentoring the team developed their own internal ERP entirely themselves — and today they run, operate, and develop it on their own.
We had an external quote on the table for a new ERP at around CZK 2 million. Under Jan's mentoring we built our own internal ERP completely on our own. Today we run, operate, and continue to develop it ourselves — for a fraction of the originally offered price.
My Aura
A distributor of cleaning automation and robots with 20 years in the market and 2465+ customers. Under my mentoring the team developed both the website and an internal operations ERP themselves — better adapted to their real business needs.
We had a quote of around CZK 150,000 for an external website. Jan guided us through the whole process and under his mentoring we built both the site and the internal operations ERP ourselves. The resulting system is much better adapted to our real business needs and we can extend it flexibly on our own.
A technical deep-dive into building an autonomous, self-repairing HTML parser pipeline using three parallel LLM witnesses, arbitration, and real-time semantic verification.
A production residential proxy pool — nine LTE modems, two Raspberry Pis, Playwright as the rotation engine, and the IP economics that make it work.
I take on selected projects and I'm open to a long-term role. I enjoy AI systems, inference, data infrastructure, and integrations the most. I work from Prague or remotely, and email is the fastest way to reach me.