Hi HN! I wanted to test the custom skills I've been building for Claude (https://github.com/josh-gree/gen-art-framework – see PRs/issues for the workflow patterns), so I asked it to build an automated generative art gallery from scratch.
Result: https://josh-gree.github.io/gen-art-gallery/
Repo: https://github.com/josh-gree/gen-art-gallery
Everything here was done through conversation with Claude:
- Set up the repo structure and GitHub Actions workflow
- Built the static gallery site (HTML/CSS/JS)
- Debugged workflow failures (Python version issues, uv configuration)
- Added incremental generation (only regenerate changed scripts)
- Added UI features (modal view, source links)
- Created example art scripts
The skills handle the full development lifecycle:
- Creating tickets to capture intent (not implementation plans)
- Planning implementations for existing tickets
- Executing on plans
- Reviewing PRs and addressing feedback
- Managing git worktrees for parallel development
The goal wasn't the gallery itself, but pinning down a good set of patterns for how Claude should handle planning, implementing, and reviewing changes in a real project. The PRs and issues in the gen-art-framework repo show the iteration on these workflows.
Curious if others are building similar "meta" workflows for AI coding assistants – patterns that work well across different projects?
*Summary:*
Senior Software/Data Engineer with a strong mathematical and computational modelling background. I build high-reliability data systems, complex ETL/ELT pipelines, and ML-ready data platforms—especially where datasets are large, irregular, hierarchical, or scientifically complex.
Most recently, I’ve been designing and operating large-scale data infrastructure for high-dimensional biological datasets (100k+ samples), unifying heterogeneous storage formats into lineage-aware catalogues, creating ontologies for hierarchical labels, building QC pipelines in Dagster, developing synthetic single-cell data generators, and working closely with domain scientists to formalise and scale experimental and computational workflows.
Previously: large-scale mobile-network analytics for humanitarian agencies; climate/energy data engineering; ad-tech pipelines; and HPC-driven modelling from computational research.
I’m looking for roles where difficult data problems, scientific or ML-adjacent pipelines, or complex modelling workflows need to be made robust, reproducible, and scalable. Prefer small teams, high ownership, and work with real impact.
*What I offer:*
– Architecture & implementation of reliable data/ML platforms
– Workflow orchestration, data governance, and reproducibility
