Skills

My work sits at the intersection of physics, AI, and infrastructure. I enjoy combining rigorous experimentation with pragmatic engineering so that research teams can ship trustworthy results without losing speed.

Programming & Data Science

Python & PyData

95%

7+ years Daily driver for experimentation with NumPy, pandas, SciPy, Plotly, and friends.

PyTorch

75%

4+ years Research-grade models, including differentiable physics workflows.

Probabilistic modelling

25%

1+ years Bayesian inference, density estimation, and uncertainty-aware ML for scientific data.

Python Ecosystem & Packaging

Isolated environments

95%

7+ years venv, Conda, uv, and Poetry to keep research stacks reproducible across machines.

Packaging & publishing

75%

5+ years PEP 517 builds,linting (ruff) ,wheels, versioning, and PyPI/internal releases with automated checks.

Scientific Computing & HPC

Containerised workflows

95%

4+ years Docker/Podman, Singularity/Apptainer, buildx caching, and registry hygiene for portable research pipelines.

GPU workflows

95%

4+ years CUDA images, multi-GPU scheduling

Slurm & Lmod stacks

85%

4+ years Scheduling, accounting, and user enablement for institute-scale clusters.

Reproducibility & Research Software

Git + testing

85%

8+ years Git, pytest, mypy, and packaging that link notebooks to published figures.

Documentation pipelines

75%

7+ years Sphinx/Markdown workflows with consistent templating so collaborators can follow every experimental step.

Scientific writing (LaTeX)

70%

7+ years Scientific manuscripts and thesis with LaTeX classes, BibTeX/Biblatex, Zotero/Citavi

DevOps & Platforms

Security & monitoring

70%

1+ years CVE tracking, dependency scanning, and lightweight observability for HPC services.

Cybersecurity & hardening

70%

2+ years CVE triage for package stacks, kernel/userspace hardening baselines, and coordinating mitigations.

GitLab/GitHub automation

75%

5+ years CI/CD, container registries, and release workflows for research services.

GitOps workflows

65%

1 years Declarative deployment patterns, repository-driven operations, and reviewable infrastructure changes.

LLM Applications & Agentic AI

Local and served LLMs

70%

3+ years Ollama and vLLM for local experimentation, inference services, and controlled model deployments.

RAG systems

65%

2+ years Retrieval-augmented generation patterns that connect language models with project, document, and domain knowledge.

Agentic AI tooling

100%

1+ years Practical use of Claude Code, Codex, Opencode, and MCP for codebase navigation, implementation support, and review workflows.