Hugo Math, PhD

Senior Agentic AI Engineer @ BMW Group · PhD candidate


Research

🔬 I am a researcher and engineer at the intersection of causal machine learning, foundation models, and agentic AI systems. I currently work as a Senior Agentic AI Engineer at BMW Group (Munich), where I design and deploy large-scale multi-agent systems for predictive maintenance and supplier intelligence.

My PhD (University of Augsburg, supervised by Prof. Dr. Rainer Lienhart, defended my PhD thesis in October 2026) focuses on scalable causal discovery and sequence modeling in high-dimensional discrete event data.

I have authored 8 publications across AAAI, NeurIPS, ICLR, ICML, filed 6 patents at the German patent offices, and serve as a reviewer at NeurIPS and ICML 2026.

⭐️ Open Source:

  • 📦 seq2cause: A Python package for population and sample-level causal discovery in high-dimensional event sequences via neural autoregressive density estimation.

💻 Stack: PyTorch, Transformers, LLMs, Causal Discovery, Graph Representation Learning.


Research Interests

  • Causal Discovery from Observational Data
  • Discrete Event Sequence Modeling
  • Foundation Models
  • Causal Foundation Models / In-Context Causal Discovery
  • Multi-Agent Agentic Systems
  • Predictive Maintenance & Industrial AI
  • Scalable GPU-Parallelized Graph Learning

News

  • 📢 [2026]: Our paper DES-JEPA: Learning a Shared Vocabulary in High-Dimensional Discrete Event Sequences was accepted to the NeurIPS 2026 Workshop (FMTS) (with Maximilian Forstenhäusler and Christos Anagnostopoulos). I will be attending NeurIPS 2026 in Sydney, Australia!

  • 🎓 [2026] Successfully defended my PhD thesis in October 2026.
  • 🏅 [2026] Awarded ICML 2026 Gold Reviewer distinction — signed letter.
  • 📄 [2026] New preprint: Learning to Predict, Discover, and Reason in High-Dimensional Discrete Event Sequences (arXiv:2603.16313).
  • 🔧 [2026] Released seq2cause — open-source library for causal discovery from event sequences (PyPI / GitHub).
  • 📢 [2026] Paper accepted at ICML 2026 Workshop SPIGM: Your Autoregressive Model Already Reveals the Causal Graph.
  • 📢 [2026] Two papers accepted at ICLR 2026 Workshops (TSALM + LLM Reasoning).
  • 🏢 [2026] Joined BMW Group as Senior Agentic AI Engineer, deploying multi-agent systems for procurement AI.
  • 📢 [2025] Two papers accepted at NeurIPS 2025 Workshops (CauScien + SPIGM).
  • 📢 [2025] Paper accepted at AAAI 2025 main track.

Publications

(Full list on Google Scholar)

DES-JEPA: Learning a Shared Vocabulary in High-Dimensional Discrete Event Sequences
H. Math, M. Forstenhäusler, C. Anagnostopoulos · NeurIPS 2026 Workshop on Foundation Models for Temporal Systems: From Forecasting to World Modeling · [2026]

Your Autoregressive Model Already Reveals the Causal Graph
H. Math, R. Lienhart · ICML 2026 Workshop SPIGM: Structured Probabilistic Inference & Generative Models · [2026]

seq2cause: Sample- and Population-Level Causal Discovery from Event Sequences using Autoregressive Models
H. Math · Open-source library · PyPI / GitHub · [2026]

Learning to Predict, Discover, and Reason in High-Dimensional Discrete Event Sequences
H. Math · arXiv preprint arXiv:2603.16313 · [2026]

Neuro-Symbolic Rule Discovery: Empowering LLMs with Causality for Vehicle Diagnostics
H. Math, L. Julian, O. Stefan, R. Lienhart · ICLR 2026 Workshop on Logical Reasoning of Large Language Models · [2026]

Context-Informed Sequence Classification: A Multimodal Approach to Vehicle Diagnostics
H. Math, R. Lienhart · ICLR 2026 Workshop on Time Series in the Age of Large Models (TSALM) · [2026]

One-Shot Multi-Label Causal Discovery in High-Dimensional Event Sequences
H. Math, R. Schön, R. Lienhart · NeurIPS 2025 Workshop CauScien: Uncovering Causality in Science · [2025]

Towards Practical Multi-Label Causal Discovery in High-Dimensional Event Sequences via One-Shot Graph Aggregation
H. Math, R. Lienhart · NeurIPS 2025 Workshop SPIGM: Structured Probabilistic Inference & Generative Models · [2025]

Harnessing Event Sensory Data for Error Pattern Prediction in Vehicles: A Language Model Approach
H. Math, R. Lienhart, R. Schön · The 39th Annual AAAI Conference on Artificial Intelligence (AAAI 2025) · [2025]


Beyond Research

  • 💡 Industry: Freelance computer vision engineer — automated semiconductor wafer inspection systems.
  • 🇸🇪 Exchange: Chalmers University of Technology, Gothenburg, Sweden.

I am actively exploring research scientist roles at labs working on causal temporal reasoning and event sequence foundation models. Feel free to reach out at hugo.mathh@gmail.com.

I’m always happy to chat about research, causal AI, or agentic systems — reach out on LinkedIn or by email.