Highlighted Projects

WISPER – AI-Based Language Support for Kindergarten Children

March 2025 – August 2026

WISPER (Wiener Spracherfahrung) is a child-friendly, AI-based language learning experience developed by the AIT Center for Technology Experience with MA 10 – Stadt Wien Kindergärten. Children use a playful telephone interface to create adaptive stories at their own pace; WISPER understands multiple languages, responds in German, and prints a picture from each story.

qwen3.8-27b-agentic-nvfp4

An NVFP4 checkpoint of Qwen3.8-27B for agentic and multimodal applications, combining text, image, and video understanding with configurable reasoning, structured tool calling, and up to 262,144 tokens of native context.

Voice-Interactive Portfolio Agent

June 19, 2026

A native voice AI layer for my personal website that turns a static portfolio into an active, agentic interface: visitors can speak naturally, get contextual answers, and navigate work without clicking around. Finished on June 19, 2026 with a focused one-week overall build timeline.

AI Agent Harness: Build the Car, Not Just the Engine

May 1, 2026

An applied AI and HCI article explaining why reliable AI agents need a production harness around the model: orchestration loops, context compaction, tool and skill registries, memory, lifecycle hooks, subagents, observability, and human approvals.

MedGemma FHIR-Bridge 💓

Built for The MedGemma Impact Challenge (Jan 22, 2026 - Feb 24, 2026): an on-premise self-healing pipeline that converts medical dark data into compliant HL7 FHIR R4 resources with a Verify & Repair safety loop.

Sovereign AI stack from Level 0 to Level 8

A production-grade, self-hosted "Sovereign AI" stack that evolves local LLMs into a complete platform with integrated RAG, agentic orchestration, and full-stack observability.

BuildsaVeR: A Virtual Reality Experience for Construction Workers safety training

buildsaVeR is designed to supplement safety training in the future and raise awareness of safe working practices. The biggest advantage: mistakes are completely harmless here - we promise!

Gaze, Gesture and Home Automation

An innovative system that combines eye-tracking, gesture recognition, and home automation to create a seamless smart home experience.

TABLE UI - TAngiBLE User Interface

A tangible user interface that transforms ordinary tables into interactive surfaces, enabling new forms of physical-digital interaction.

EnthusiastiCan - Enthusiastic Trash Can

An interactive trash can that uses computer vision and gamification to encourage proper waste disposal and recycling.

Voice Controlled Robotic Arm

A robotic arm that can be controlled through natural voice commands, making robot manipulation more intuitive and accessible.

Multi-functional Mobile Robot

A versatile mobile robot platform featuring object tracking, leader-follower behavior, and tag-based navigation capabilities.

Vision-based Sorting System

An industrial sorting system using computer vision and a robotic manipulator for automated object recognition and sorting, designed for Industry 4.0 applications.

Navigation & Path Planning for Mobile Robot

Developed advanced navigation, path planning, and motion planning capabilities for mobile robots including ARTI Chasi and Franka Panda.

EMG-Based Control of a 5 DOF Robotic Manipulator

Designed a system to control a 5-degree-of-freedom robotic manipulator using electromyography (EMG) signals for intuitive human-robot interaction.