Dr. rer. nat. Momme Allalen
Computational Physicist · HPC & AI Consultant
I help scientists run faster — from quantum spin systems to extreme-scale simulations and AI workloads on some of Europe's largest supercomputers at the Leibniz Supercomputing Centre (LRZ), Bavarian Academy of Sciences and Humanities, Garching near Munich.
01 · About
Making science compute faster.
I work full-time in the High-Performance Computing group at LRZ, specializing in high-level application support, training, and extreme-scale computing. My daily work: taking scientific codes from every discipline — CFD, physics, chemistry, earth sciences, life sciences — and making them fly on modern heterogeneous CPU + GPU architectures.
Trained as a computational physicist (Dr. rer. nat., Universität Osnabrück) with molecular dynamics and Monte Carlo roots, I've spent my career at the intersection of science, algorithms and hardware — from quantum spin systems to petascale fluid dynamics, and today to AI-accelerated research workflows on NVIDIA DGX systems and Europe's emerging exascale & AI-factory infrastructure.
As an NVIDIA DLI Certified Instructor, I teach CUDA fundamentals and advanced multi-GPU scaling of CUDA C++ applications across many nodes — and I train researchers across Europe on interactive AI workflows on the LRZ AI Systems.
Application scaling & support
High-level support for LRZ users: porting, parallelizing and tuning scientific codes on novel HPC architectures.
Training & teaching
DLI CUDA courses, EuroHPC training events, GPU & extreme-scaling workshops for the research community.
Benchmarking & evaluation
Evaluating high-end supercomputers — from the Mont-Blanc ARM prototype to SuperMUC generations and AI systems.
Software stacks & tooling
Building and maintaining scientific software for current & upcoming architectures with Spack, containers and CI.
02 · Expertise
What I bring to your code.
Two decades of parallel computing — from message passing on clusters to multi-node GPU training runs. The tools change; the physics of performance doesn't.
Performance Engineering
Low-level optimization, performance analysis and tuning on modern architectures: profiling, vectorization, memory hierarchy, node-level bottlenecks.
Parallel Programming Models
Choosing and mastering the right model for the machine — and keeping codes performance-portable across vendors and generations.
GPU & Heterogeneous Computing
Porting and optimizing software on CPU+GPU supercomputers; advanced multi-GPU scaling of CUDA C++ applications across multiple nodes.
AI & Deep Learning Infrastructure
User support for ML/DL on LRZ's NVIDIA DGX systems: interactive Jupyter workflows, environment management, and scaling training to many GPUs.
Scientific Software Stacks
Installing and maintaining research software for current & upcoming processor architectures — reproducible, containerized, automated.
Benchmarking & System Evaluation
Evaluating the performance of high-end supercomputers and novel architectures — including ARM-based HPC prototypes and many-core systems.
03 · AI & Modern Workflows
What changed since 2011 — and how I work now.
This page once ended with a “under construction” GIF. The science moved on: today my workflows combine classical HPC craftsmanship with AI tooling — LLM-assisted code engineering, GPU-native machine learning, and fully reproducible environments.
AI-assisted code engineering
Using LLM copilots to accelerate porting and refactoring of legacy scientific codes (Fortran/C++ → CUDA/SYCL), generate tests and documentation — every suggestion validated by profiling and correctness checks, never blind trust.
ML & DL at scale on LRZ AI Systems
Supporting researchers on NVIDIA DGX: interactive terminal & Jupyter workflows, from single-GPU prototyping to distributed multi-node training — the same material I teach in EuroHPC training events.
Reproducible & automated environments
Containerized scientific software (Apptainer, Charliecloud) deployed on HPC systems, Spack environments for emerging architectures, CI pipelines that keep research codes building — a workflow we published at SC21.
Quantum simulation meets AI & HPC
Tensor-network quantum circuit simulation at exascale, containerized quantum software (QuantEx), and benchmarking emerging computing paradigms — where my physics roots meet tomorrow's hardware.
The machines behind the work @ LRZ
04 · Journey
Two decades of scaling science.
Every hardware generation brought a new programming paradigm — and a new chance to make research codes faster.
HPC & AI Consultant · Extreme-Scale Support
Leibniz Supercomputing Centre (LRZ), Garching
High-level application support and extreme scaling on CPU+GPU systems; NVIDIA DLI Certified Instructor for CUDA & multi-node multi-GPU techniques; trainer in EuroHPC (MultiXscale) events on interactive AI workflows on the LRZ AI Systems; AI-assisted performance engineering and containerized, reproducible software stacks.
Exascale & Quantum Simulation Software
SC21 — three works at the supercomputing flagship conference
Tensor-network quantum circuit simulation at exascale; deploying containerized QuantEx quantum simulation software on HPC systems; a next-generation discontinuous Galerkin solver for high-resolution lung airflow simulations.
Petascale Performance & Portability
SuperMUC Phase 2 · PASC · HPCS · ISC · SC16
Extreme scale-out of applications on SuperMUC Phase 2; optimization of the parallel CFD code MGLET across HPC platforms (PASC'19); node-level optimization with performance portability (HPCS'19); fast matrix-free discontinuous Galerkin kernels (ISC'17); the Mont-Blanc ARM-based prototype for future HPC systems (SC'16).
The Many-Core Frontier
Intel Xeon Phi · SuperMIC · PRACE
First experiences with the Intel MIC architecture at LRZ; porting a real-world seismological application to Intel MIC (PRACE whitepaper); organizing the Intel MIC & GPU programming workshop series and extreme-scaling workshops at LRZ.
Dr. rer. nat. — Computational Physics
Universität Osnabrück · then LRZ HPC group
Doctorate in computational physics: molecular dynamics and Monte Carlo simulations of macroscopic systems and quantum theory; graduate research on surfaces and interfaces assembled from clusters and molecules. The beginning of a lifelong obsession: making physics computable.
05 · Publications
Selected work.
A few highlights — the full archive of publications, talks and posters lives on the pages below.
A next-generation discontinuous Galerkin fluid dynamics solver with application to high-resolution lung airflow simulations
Kronbichler, Fehn, Munch, … Allalen, Schulz, Wall — SC '21, Int. Conference for High Performance Computing, Networking, Storage and Analysis.
ACM DL ↗Tensor Network Circuit Simulation at Exascale
Brennan, Allalen, Brayford, Hanley, Iapichino, O'Riordan, Doyle, Moran — 2nd Int. Workshop on Quantum Computing Software, SC21.
arXiv ↗Deploying Containerized QuantEx Quantum Simulation Software on HPC Systems
Brennan, Allalen, Brayford, Hanley, Iapichino, O'Riordan, Moran — 3rd Int. Workshop on Containers and New Orchestration Paradigms (CANOPIE-HPC), SC21.
arXiv ↗Performance Optimisation of the Parallel CFD Code MGLET across Different HPC Platforms
Sakai, Mendez, Strandenes, Ohlerich, Pasichnyk, Allalen, Manhart — Platform for Advanced Scientific Computing Conference, Zurich.
ACM DL ↗Fast Matrix-Free Discontinuous Galerkin Kernels on Modern Computer Architectures
Kronbichler, Kormann, Pasichnyk, Allalen — ISC 2017 High Performance Computing, LNCS 10266, Springer.
Springer ↗Extreme Scale-out SuperMUC Phase 2 — Lessons Learned
Hammer, Jamitzky, Satzger, Allalen, et al. — ParCo 2015, MAXI mini-symposium on multi-system application extreme-scaling. See also: The Mont-Blanc Prototype (SC16).
arXiv ↗06 · Teaching & Training
Passing the tools forward.
Training is half the job: the fastest code is the one a researcher can write themselves.
CUDA & Multi-GPU Scaling
Certified instructor for Deep Learning Institute courses: fundamentals of CUDA C/C++ and advanced techniques for scaling CUDA C++ applications across multiple GPUs and multiple nodes.
AI Workflows on LRZ AI Systems
Trainer in European HPC training events: working interactively with the LRZ AI Systems via terminal and Jupyter notebooks — from single-GPU workloads to distributed deep learning.
GPU & Extreme-Scale Workshops
Co-organizer and instructor of the Intel MIC & GPU programming workshop series, extreme-scaling workshops at LRZ, and university teaching on parallel programming.
Teaching archive →Links & resources I keep coming back to
07 · Contact
Boltzmannstr. 1 · 85748 Garching bei München · Germany
Let's scale your science.
Application scaling, GPU porting, training, or a collaboration on AI-driven simulation — my inbox is open.