Projects

Research, development, and open-source work

Optimizing I/O Performance of Scalable ML Workflows in HPC Systems

Optimizing I/O Performance of Scalable ML Workflows in HPC Systems

Master ThesisHPCDALISLURM

M.Sc. thesis investigating data-loading bottlenecks in large-scale deep learning workflows. Benchmarked NVIDIA DALI, TFRecords, and WebDataset across PyTorch and TensorFlow on SLURM-managed HPC clusters, producing reproducible I/O performance analyses.

TensorRigs.com

TensorRigs.com

AstroML HardwareDeep Learning

AI Hardware and ML GPU resource site for deep learning practitioners. Expert guides for GPU comparisons (NVIDIA RTX, A100, H100), deep learning system selection, and PyTorch/TensorFlow hardware optimisation. Built with Astro, deployed on Netlify with Cloudflare.

TensorDocs — HPC and ML Technical Documentation

TensorDocs — HPC and ML Technical Documentation

StarlightHPCDocumentation

Comprehensive documentation site built with the Starlight framework covering HPC workflows, NVIDIA DALI tutorials, GPU acceleration best practices, and deep learning optimisation techniques.

Starlight Documentation: Large-Scale ML in HPC

Starlight Documentation: Large-Scale ML in HPC

DocumentationHPCDeep Learning

Additional documentation from thesis work covering software environment setup, dependency management, and common challenges in HPC deep learning workloads.

Sentiment Analysis on Course Feedback

Sentiment Analysis on Course Feedback

NLPPythonSentiment Analysis

NLP project applying sentiment analysis to student feedback from a university seminar. Covers text preprocessing, model evaluation, and review of relevant scientific papers.

University Research Group Lab Website

University Research Group Lab Website

Web DevelopmentResearchUniversity

Research group website for the AI and Swarm Intelligence Lab at Göttingen University, supporting work in Swarm Intelligence, Wireless Sensor Networks, and Artificial Intelligence.

ZombieSim — Parallel HPC Infection Simulator

ZombieSim — Parallel HPC Infection Simulator

C++OpenMPHPCParallel Computing

Multithreaded C++ infection simulator built for the GWDG HPC Practical Course. Combines a configurable simulation engine with visualisation, demonstrating OpenMP parallelisation on HPC cluster infrastructure.

Self-Hosted GDPR-Compliant Infrastructure Stack

HetznerCoolifyDockerGDPRSelf-Hosted

Production self-hosted stack on a Hetzner VPS managed via Coolify, replacing commercial SaaS tools with open-source alternatives: n8n, Plausible Analytics, Twenty CRM, Listmonk, NocoDB, Vaultwarden, and more. All GDPR-compliant, zero third-party data sharing.

Self-Hosted Infrastructure

GDPR-compliant open-source stack running on a Hetzner VPS, managed via Coolify

n8n

Replaces Zapier, Make

Workflow automation and AI pipeline orchestration.

Plausible Analytics

Replaces Google Analytics

Privacy-first, cookieless web analytics. No consent banner needed.

Twenty CRM

Replaces Salesforce, HubSpot

Open-source CRM for managing contacts and client relationships.

Listmonk

Replaces Mailchimp, ConvertKit

Self-hosted email newsletter and mailing list manager.

NocoDB

Replaces Airtable

No-code database UI on top of existing databases.

Uptime Kuma

Replaces Pingdom, UptimeRobot

Self-hosted uptime and status monitoring for all services.

Vaultwarden

Replaces LastPass, 1Password

Self-hosted Bitwarden-compatible password manager.

Rybbit

Session replay and product analytics

Open-source session recording and user behaviour analytics.

Hosting Stack

Hetzner VPS Coolify Docker Cloudflare Postale CI/CD