About

What I've actually built, and where I haven't.

I'm John DeLeon. 14 years in enterprise software, most of it in the insurance industry, writing C# microservices for carrier and web integrations — promoted from developer to Lead Software Developer to Software Engineer along the way. For 5 of those years I was the Java technical lead directing an event-driven claims-system implementation alongside the software provider.

I finished an MS in Computer Science at Angelo State University in December 2025, with an ML focus and Python coursework. That degree, not a bootcamp or a weekend course, is where the AI/ML side of this started.

The throughline

14 years of enterprise experience, now deliberately expanding into AI and business automation. Not a pivot away from what I know — an extension of it. That's what this site is: the record of that extension, including what I still don't know.

What I've built

  • A DOD-funded capstone— a Kafka/Spark/Neo4j/TimescaleDB pipeline training graph neural network models for anomaly detection with explainability, run against a real IEEE Dataport hardware-in-the-loop testbed dataset. Five-person team, DOD/ARL grant, published paper. A class project, not a production deployment.
  • A fraud detector for job postings— text classification with a full backend/frontend and an AWS CDK deployment around it. Three-person class project, not a solo build. The trained model hits 98% accuracy offline; the deployed demo doesn't call it yet — that gap is the actual story.
  • A production ordering site for a real cottage-food bakery (Astro, Supabase, Cloudflare) — I directed an AI-assisted build with Claude Code, meaning I prompted, reviewed, and evaluated its output rather than hand-writing every line, and configured the self-hosted n8n automation behind it. This one is live and in use, not a demo.

What I don't have standing on

Two things I won't pretend expertise in just because I touch them:

  • Performance engineering and security. I think about them when I build. I don't have deep expertise in either, and I won't write like I do.
  • Frontend design. I'm mid at best. I lean on AI tooling to do the actual design work, and I'll say so when it's relevant.

If you want the field notes, they're on the blog. If you're evaluating whether to hire me for automation or AI-integration work, that's a separate conversation — see what I work on.