A selection of projects from my professional experience and academic work — ranging from AI orchestration platforms to embedded systems.
AI / ML03/2025 – 08/2025
POKIO – Prompt-Optimized AI Orchestration
Prim-Agency GmbH (Bachelor Thesis)
An orchestration platform that generates complete frontend applications from natural language prompts. Investigated multi-agent vs. single-agent orchestration strategies for React-based frontend development.
Autonomous generation of full React frontend applications from prompts
Comparison of multi-agent vs. single-agent orchestration approaches
Integration of LLM pipelines with structured output validation
A local AI agent that runs a small LLM on-device. Features a central API server communicating with the LLM, a custom Signal-MCP server, voice interaction, and integrations with VDB, DB, MCP, VTS, and TTS.
Local LLM execution for privacy-first AI interactions
Custom Signal-MCP server for messaging integration
Voice control via speech-to-text and text-to-speech
Performance-optimized with WebSockets and local networking
A full-stack application built on top of POKIO. Enables creation of complete frontend apps from prompts with enterprise-grade authentication, multi-tenancy, and real-time streaming.
Multi-tenancy scope with isolated data per tenant
Keycloak authentication with JWT token exchange
HTTP-only cookies for enhanced security
Server-Sent Events (SSE) for real-time streaming
NGINX as reverse proxy for the React frontend
ReactTypeScriptFastAPIKeycloakSSENGINXDocker
Web App05/2024 – 03/2025
Niko-Niko – Team Mood Tracker
Novatec Consulting GmbH
Maintenance and further development of an internal team mood tracking application. Acted as SCRUM Master, leading sprint ceremonies and monitoring performance via Azure Portal.
Bug identification and resolution across the full stack
Implementation of new features based on team feedback
SCRUM Master: sprint planning, reviews, and retrospectives
Performance monitoring with Azure Portal & Application Insights
ReactTypeScriptAzureApplication InsightsSCRUM
Embedded10/2022 – 02/2023
Autonomous Parking Vehicle
Hochschule Darmstadt
A sensor-equipped model vehicle capable of autonomously detecting and using parking spaces. Included an Android app for remote control and communication with the vehicle.
Sensor-based autonomous parking space detection
Android app for remote control and live communication
Efficient communication interfaces between app and vehicle
Embedded systems programming for real-time control