My Portfolio
Research, Development, and Innovation
Monti Platform (CONAN Study)
University Hospital Düsseldorf | 2022-2025
Development of an AI-based platform for continuous monitoring of heart patients using wearables. The system uses self-supervised contrastive learning for early detection of decompensation events in heart failure patients.
- 290 patient prospective cohort study
- 90 days continuous monitoring
- Deep learning for anomaly detection in vital signs
- Real-time processing of wearable data
BSI IT-Grundschutz AI Search
2025
Development of an intelligent search system for BSI IT-Grundschutz documentation. The system provides precise answers to compliance questions with exact source citations.
- RAG (Retrieval-Augmented Generation) architecture
- 1000+ pages searchable
- 95% time savings for compliance queries
- Vector-based document search
Technology:
Python, Docker, ChromaDB, FastAPI
Bachelor Thesis: Active Pre-Training with Phasic Policy Gradient
2023
This thesis combines Active Pre-Training (APT) with Phasic Policy Gradient (PPG) for on-policy reinforcement learning. I addressed the Parallelism-Variance Dilemma in contrastive learning when using multiple parallel environments.
- SimCLR objective for encoder training
- Particle-based entropy rewards
- 62.5M pre-training steps
- Stabilization techniques: gradient clipping, entropy regularization
TurtleBot Autonomous Navigation
2021
Development of autonomous navigation technologies using the TurtleBot. The robot serves as a test platform for reinforcement learning algorithms and for optimizing control systems.
- Sim-to-Real Transfer
- ROS (Robot Operating System)
- Deep Reinforcement Learning
- Sensor fusion and environment mapping
Interested in Working Together?
Let's discuss your next AI project.
or email me: [email protected]