I build machine-learning systems and iOS apps - lately, on-device vision that reads handwriting and turns paper into data. Off the clock I shoot the cities, mountains, and trips I find myself in.
A field CRM built end to end in SwiftUI on Supabase. An on-device vision pipeline - Qwen 3 8B running through MLX with Whisper for dictation - reads handwritten service records straight off the page and turns them into structured, searchable history. No round trips to a server, no manual retyping.
A custom booking and scheduling system for a rehabilitation and medical clinic, designed and shipped independently. Built around how the staff actually book rooms, therapists, and recurring patient slots rather than a generic calendar.
Video-based estimation of physical workload using YOLO, paired with hands-on data collection in the real environment it measures. The hard part is rarely the model - it is gathering footage that reflects the actual strain people put on their bodies.
Complete stock, orders, and receipt management for a warehouse, wired directly into the accounting software so a movement on the floor and the books never drift apart. Receiving, picking, and reconciliation in one place.
A weekly reporting and optimization agent that pulls performance from BigQuery, reasons over it with Claude through MCP, and proposes and applies account changes on its own. The manual Monday-morning ads ritual, handed to a tool that does it every week.
I have played ice hockey for 17 years, and still do - currently in the 2. liga with Nový Jičín and BO Ostrava. The same thing pulls me to a rink and to a hard engineering problem: a clear goal, a team that has to read each other, and the small margins that decide it.
From Swift on the device to SQL and agents on the back end - whatever the problem actually needs.