// projects

Things I've shipped, end to end

Not case studies of work I helped with — products I designed, built, deployed, and pay the hosting bill for.

PRODUCT // LIVE

DocFetch

An AI document collection platform. You tell it what you need and who has it — it drafts the outreach, tracks what's missing, sends reminders, and delivers an organized packet.

Planner, collector, and isolated classifier agents with bounded execution loops and restricted tool access — designed to reduce prompt-injection risk
Persistent workflow orchestration via PostgreSQL state transitions, driven by AgentMail webhooks — no polling
Stripe Checkout, subscription billing, waitlists, authentication, and privacy-first automatic document deletion
Built for loan officers, bookkeepers, closing teams, and property managers chasing paperwork by hand

Python · FastHTML · Anthropic · Supabase / PostgreSQL · AgentMail · Stripe · Fly.io

docfetch.tech ↗

OPEN SOURCE // MIT LICENSE

Aquifer

A self-hosted Go runtime built for the moment an agent workload stops being a toy: durable queuing, controlled dispatch, and backpressure so a traffic spike doesn't take down the backend it's calling.

Durable job queue with SQLite persistence and crash recovery — no external database required
Per-upstream rate limiting and dynamic pacing driven by the upstream’s own response headers
HTTP API and MCP stdio adapters, so it drops into existing agent/tool-calling stacks
L8 protocol for trustless webhook delivery using public-key crypto instead of shared secrets

Go · SQLite · MCP · Server-Sent Events

GitHub — rjpruitt16/aquifer ↗

// building something similar?

If you're picturing your own version of this

Most of what's above started as someone describing a problem out loud. If you've got an AI product idea and need someone who can both design it and actually build it, that's the conversation I have on the consulting page.

Work with me →