Curriculum Vitae
Andreas Hermann
AI safety researcher and ML engineer. PhD (magna cum laude), REACH Award. 7+ years across research and production machine learning. Download full CV (PDF) →
Experience
-
2026 – present
Independent AI Safety Researcher
BlueDot Impact transition grant (from Aug 2026)
Full-time self-directed research — open-weight model safety, compositional misalignment, interpretability and evaluation.
-
2026
AI Safety Research Fellow
Safe AI Germany (SAIGE)
Inoculation against model poisoning — data-level defenses against emergent misalignment.
-
2025 – 2026
AI Alignment Research Fellow
AI Safety Global Society (ARF) · ENAIS network
Inaugural ARENA-based cohort. Mechanistic interpretability, scalable oversight, agent foundations, robustness. Published a peer-reviewed paper on multi-agent alignment (HCII).
-
2024 – 2026
Director of Data & AI/ML Engineering
Enterprise product-data SaaS · Cologne, Germany
Led in-house ML services (~€130k/yr combined savings & revenue). Architected and deployed deep-learning, multimodal, and fine-tuned (LoRA/PEFT) models; matured MLOps for 5+ production models.
-
2023 – 2024
Data Scientist
Global product-content network · Cologne, Germany
Built and personally deployed 4+ ML models to production using transformers and multimodal processing on GCP (Vertex AI, BigQuery, Dataflow).
-
2018 – 2022
Research Assistant
University of Münster · Chair for Information Systems
Led a ~€300k EU-funded research project (10 reports, 6 countries). Supervised data-science seminars. Published 17 peer-reviewed articles (250+ citations at the time).
Education
-
2018 – 2022
PhD, Information Systems
University of Münster — magna cum laude
REACH Thesis Award 2022.
-
2015 – 2017
MSc, Information Systems
University of Münster — grade 1.3
CLAAS AlumniUM Master Award 2018. Exchange semester, Karlstad University (Sweden), 2016.
-
2012 – 2015
BSc, Information Systems
University of Münster — grade 1.8
Beta Gamma Sigma membership.
AI safety training
- Technical AI Safety — BlueDot Impact (2026)
- Alignment Research Fellowship — AI Safety Global Society (2025)
- AI Safety Fundamentals: Alignment & Intro to Transformative AI — BlueDot Impact (2024–25)
- ARENA curriculum · Deep Learning & ML Specializations — DeepLearning.AI / Stanford
Selected skills
ML & research: PyTorch, TensorFlow, transformers, fine-tuning (LoRA/PEFT), multimodal models, inference optimization (quantization), mechanistic interpretability, steering vectors, evaluations.
Engineering & MLOps: Python, SQL, Docker, GCP (Vertex AI, BigQuery, Cloud Run), MLflow / Metaflow, Elasticsearch, ETL & data pipelines.
Languages: German (native), English (professional).