RG26

51.47° N / 7.25° E

René GlitzaNerd with ashirt.

Researcher. Builder. Founder.

I build AI systems that learn from distributed, private, real-world data.

01 / Signal

Between the model and the world.

I work where machine-learning research meets systems engineering and industrial reality. My focus is not only the model, but the path that makes it reproducible, distributed, and useful.

That path has taken me from acoustic condition monitoring and federated learning to open-source software, startup building, electronics, measurement systems, and the infrastructure that lets experiments become dependable practice.

Off hours

  • Espresso
  • Sailing
  • Open-source smart home
Portrait of René Glitza
Research / systems / industrial practice

02 / Practice

One chain. Three contexts.

  1. 01

    Adaptive and personalized Federated Learning, reinforcement learning, heterogeneous and limited-label data, anomaly detection, and acoustic condition monitoring.

  2. 02

    Co-Founder

    NexuFed AI

    Collaborative, private Industrial AI: systems that learn across organizations without requiring their raw data to be centralized.

  3. 03

    End-to-end systems engineering

    AI-Gruppe

    Hands-on implementation and technical leadership across PCB design, firmware, software, data science, and simulation, taking three products from proof of concept to international market launch.

  1. 01

    Research question

  2. 02

    Reusable ML system

  3. 03

    Kubernetes training infrastructure

  4. 04

    Industrial application

I build reproducible MLOps workflows and Kubernetes-based AI-training clusters that connect distributed workloads, data paths, and experiment tracking to the systems they must eventually serve.

03 / Selected systems

Work that travels.

  1. 01Open-source ML systems

    NexuML

    A modular PyTorch framework for composable machine-learning pipelines, typed reusable components, and reproducible experiments.

    Explore NexuML on GitHub ↗
  2. 02Federated-learning infrastructure

    NexuFL

    Adaptive infrastructure for distributed, private learning across research and production environments.

    Visit NexuFed AI ↗
  3. 03ICASSP 2026 reproduction

    pFedMARL

    TD3-based cooperative multi-agent reinforcement learning for adaptive aggregation with non-IID DCASE Task 2 data.

    Explore pFedMARL on GitHub ↗
  4. 04Acoustic research data

    ASN Database

    A public database of simulated room impulse responses for acoustic sensor networks in complex multi-source environments.

    Explore ASN Database on GitHub ↗

04 / Questions

Learning under real constraints.

How do systems learn when data cannot move freely, labels are scarce, environments differ, and deployment is more than a benchmark?

  • Adaptive and personalized Federated Learning
  • Reinforcement Learning for distributed systems
  • Industrial AI and acoustic condition monitoring
  • Edge AI with heterogeneous, limited-label data
  • Reproducible MLOps and Kubernetes AI-training clusters

05 / Community

Systems scale through people.

Vice Chairman

open Skunkforce e.V.

An open technical community creating room for exchange, experimentation, and projects that benefit from collective effort.

Young Professionals Representative

VDE Rhein-Ruhr e.V.

Connecting students and early-career engineers through regional exchange, technical events, and the wider VDE Young Net.

06 / Continue

Bring a hard problem.

I am interested in collaborations where research quality, infrastructure, and practical constraints all matter.