LOG_ENTRY // MEMPHIS, TN

I build the systems
underneath AI —
then the agent on top.

7+ years shipping production software at Amazon Prime Air, Twitch, and M&T Bank. Now building AI agent infrastructure and custom AI products for teams who'd rather ship than firefight.

Rahmi Pruitt FIG. 01 — R.PRUITT
ROLE_01

Software Engineer

Distributed systems and cloud infra at Amazon Prime Air, Twitch, and M&T Bank. Go, Python, TypeScript, AWS.

ROLE_02

Founder

Built and shipped DocFetch, a production multi-agent AI SaaS, and Aquifer, an open-source agent runtime.

ROLE_03

AI Consultant

I help non-technical founders and teams turn an idea for an AI product into a working system — and a real quote.

// featured work

Two ways I've shipped this myself

PRODUCT // LIVE

DocFetch

AI document collection platform. Tell it what you need and who has it — it handles outreach, reminders, and delivery.

  • Planner, collector, and isolated classifier agents with bounded execution loops
  • Persistent workflow orchestration via AgentMail webhooks + Postgres state machine
  • Stripe billing, waitlists, auth, and privacy-first automatic document deletion
Python · FastHTML · Anthropic · Supabase · Stripe · Fly.io
OPEN SOURCE // MIT

Aquifer

A self-hosted Go runtime that stops agent retry storms before they turn into runaway LLM bills — durable queuing, paced dispatch, and cryptographic agent identity via the L8 protocol.

  • Durable job queue with SQLite persistence and crash recovery
  • Per-upstream rate limiting, backpressure, and fairness primitives
  • HTTP + MCP adapters for coordinating agentic and tool-calling workloads
Go · SQLite · MCP · Server-Sent Events

// on the record

Writing and talks on agent reliability

I write and record about the unglamorous parts of AI infrastructure — retries, rate limits, and what happens when a "simple" agent hits production traffic.

Read & watch →