Machine Learning Engineer. Physics-Informed and Structure-Aware Modeling

About

Physics-trained ML engineer working on time-series modeling, anomaly detection, and generative models with physical constraints. Experienced in leading research teams, building widely used scientific software, and translating domain knowledge into robust ML systems. Interested in data-efficient learning for dynamical systems and hybrid physics–ML approaches.

Experience

ML and Physics Research Leadership (2011–Present) Austria, France, Germany

• Led physics-driven ML projects and large scale physics projects on real-life data
• Managed €1.3M+ in competitive funding and supervised interdisciplinary teams
• Lead developer of widely used Python scientific software (SMoldeS)
• Regular invited speaker at conferences, workshops, and applied ML meetups

Selected Projects

Physics-Aware LSTM for Anomaly Classification

• Demo and code
• Achieved 57% Mahalanobis distance metric improvement for anomaly type cluster separation with physics constraints over baseline LSTM
• Built 2D diagnostic landscape separating reconstruction and physics loss for failure mode classification using PyTorch
• Documented decision framework as a 10-part public technical series reaching 50k+ impressions
• Tech Stack: pandas, PyTorch, scikit-learn, Streamlit

Agentic Framework for Music Analysis and Recommendation

• Code
• Designed constrained LLM reasoning architecture grounding every recommendation in data fields, preventing hallucination while preserving human-readable explanations
• Built hybrid system combining Spotify audio features with semantic topic modeling across 153 songs and 5 distinct musical eras
• Engineered agentic retrieval pipeline with explainability as a first-class requirement, not a post-hoc label
• Tech Stack: pandas, hdbscan, sentence transformers, PyTorch, scikit-learn, ollama, openai, Streamlit

Remaining Useful Life Prediction (NASA Turbofan Data)

• Demo and code
• Structured time-series modeling preserving unit-level dependencies
• Uncertainty-aware RUL estimates using gradient boosting and automated tuning
• Focus on evaluation aligned with maintenance decision-making

Physics-Aware Latency Prediction and Network Anomaly Detection

• Demo and code
• Physics-aware anomaly detection for networked systems
• Improved sensitivity to rare, physically meaningful anomalies while stabilizing false-positive rates
• Emphasis on interpretability and operational robustness

Talks and Community

• Invited and meetup talks on physics-informed ML
• Recent online talk on PIML for dynamical systems. Slides
• Member of Styrian vision group of the Women in AI, Austria
• Created AI-generated TedXGraz theme song; documented process in a LinkedIn article demonstrating creative application of ML in collaborative, real-world settings.

Selected Publications

Long-Lived Particle Searches at the LHC

Publication

• Python-based simulation and visualization for search optimization
• Influenced parameter choices and experimental search strategies

Snowmass 2021 Dark Showers Report

Publication

• Coordinated 50+ researchers across theory and experiment
• Shaped field-level synthesis and research directions

Constraining new physics with SModelS version 2

Publication

• Lead architect of Python codebase covering 100+ new-physics searches
• Accelerated interpretation of collider constraints on dark matter models