Solo studio ยท Since 2015

Django, Python, and the AI work that's harder than it looks.

Wildflower Software is one senior engineer, not an agency. I write the code and make the architecture calls, and I'm still around long enough to find out whether they were the right ones.

Django 2.x โ†’ 6.x upgrades LLM pipelines in production Deploys on managed Python platforms

Comfortable in Pydantic LangGraph MetaFlow Django Wagtail Celery Docker AWS Terraform

I consult

Embedded with your team for a sprint or a quarter. Architecture reviews, pairing, upgrade plans. Mostly it's making a codebase predictable enough that people stop dreading it.

How engagements work

I build

New Django apps and AI features, built end to end: schema, background jobs, evals, deploy pipeline. I write the runbook too, because I'm not going to be here forever.

Talk about a build

Notes

Notes from actual engagements.

Nov 1, 2026

What We Handed to a Framework, and What We Kept

The flower shop agents now run on PydanticAI inside Prefect flows, and their tools are still served by FastMCP. A new weekly sales note runs on its own every Monday. The allow-lists and the approval gate are still our code. Here's what changed and why.

Oct 16, 2026

Human-in-the-Wild

The most useful knowledge in a small business is the stuff the owner never writes down. Here's the pattern I'm using to get it into a recommendation engine: she texts what she sees, and the agent listens.

Oct 7, 2026

Does the Flower Shop Need an AI Gateway?

What an AI gateway does, whether the flower shop agents need one, how LiteLLM, Portkey, Cloudflare and Kong compare, and how I put gateways in for clients. For the people buying and the people building.

Read all notes โ†’

Have something gnarly?

Slow queries, a Django upgrade you keep postponing, an AI feature that needs to survive production. Tell me about it.