CV
Download my CV as PDF: 🇬🇧 English  | 🇩🇪 Deutsch
Contact Information
| Name | Niklas Schweiger |
| Professional Title | M.Sc. Student in Robotics, Cognition, Intelligence |
| niklas.schweiger@tum.de | |
| Location | Munich, Bavaria |
| Website | https://niklasschweiger.github.io |
Professional Summary
Master’s student at TU Munich specializing in machine learning, generative models, and inference-time alignment. Currently a research assistant in Prof. Daniel Cremers’ Computer Vision & AI group, working on reward-agnostic methods to steer diffusion and flow models at inference time.
Experience
-
2026 - 
Research Assistant (HiWi)
Chair for Computer Vision & Artificial Intelligence (CVAI), TU Munich
Student research assistant at Prof. Daniel Cremers’ group — the same lab where I am writing my Master’s thesis and where the ICLR 2026 paper was produced. Research spans generative models and inference-time alignment.
- Research on inference-time alignment of diffusion and flow models
- Contributed to ICLR 2026 workshop paper on trust-region noise optimization
-
2023 - 2023 
Intern – AI in Industrial Production
Siemens AG
Developed a 3D feature matching system using CNN embeddings to automate part retrieval in manufacturing databases.
- Implemented multi-view CNN pipeline for part similarity retrieval
Education
-
2023 - M.Sc.
Technical University of Munich (TUM)
Robotics, Cognition, Intelligence
- Machine Learning
- Deep Learning
- Computer Vision
- Probabilistic Graphical Models
- Exchange semester at Chalmers University of Technology, Gothenburg, Sweden
-
2020 - 2023 B.Sc.
Technical University of Munich (TUM)
Electrical Engineering and Information Technology
- Artificial Intelligence and Machine Learning (from 5th semester)
- Signal Processing
- Control Systems
-
2024 - 2025 
Awards
-
2023 
2nd Place – TUM.ai Makeathon (€2,000 prize)
TUM.ai
Built ‘Caire’, an app designed to overcome language barriers between nurses and patients. Using AI, it extracts medically relevant information from natural speech and automatically generates structured medical reports.
Publications
-
2026 
Trust-Region Noise Search for Black-Box Alignment of Diffusion and Flow Models
ReALM-GEN Workshop @ ICLR 2026
We propose a simple trust-region based search algorithm (TRS) which treats the pre-trained generative and reward models as a black-box and only optimizes the source noise. Evaluated across text-to-image, molecule, and protein design tasks. ICLR 2026, Rio de Janeiro, Brazil.