Python automation, Claude and GPT API integrations, and AI-assisted pipelines — built for real tasks, not demos. The goal is to remove repetitive work without removing judgment.
Custom Python scripts for data processing, content structuring, batch operations, API integrations, and scheduled tasks. Designed to run on your VPS, locally, or via systemd.
API key setup, prompt engineering, structured output extraction, rate limit handling, retry logic, and incremental result saving. Built to survive errors and resume gracefully.
AI-assisted schema generation for large sites. Page data in, valid JSON-LD out — validated against schema.org. Batch-capable with incremental saving.
Fetch, extract, and summarize content from multiple sources into structured JSON or formatted output. Used in the HomelessDenver.com resource verification workflow and similar projects.
Python services running as systemd units. Cron job setup. OpenLiteSpeed proxy configuration for internal APIs. I manage this stack on my own VPS and can set it up for yours.
System prompt design, few-shot examples, output format specification, and iterative testing. Most AI output quality problems are prompt quality problems.
I've been using the Claude API in production since mid-2025. Not for generating blog posts — for actual pipeline work: schema generation, research summarization, content structuring, and batch data processing.
The thing that makes AI automation reliable is the same thing that makes any automation reliable: error handling, incremental saves, sensible defaults, and a clear understanding of what you're asking for. A script that fails at item 60 of 80 and discards all progress is worse than no script at all.
I build AI integrations that work the same way on the 100th run as the first. I document them so they're maintainable after I'm done with the project.
Describe what you're doing manually and how often. I'll tell you whether it's automatable and roughly what it would take.