Solo architect. 10 production systems, 1,200+ commits, zero SaaS dependencies for core ops. I build for fun, for learning, and for businesses I'm personally involved in.
A complete operations platform built from scratch that replaced 5–6 separate SaaS products at a service business. Manages the entire job lifecycle — intake → work orders → tech queue → inspection → invoicing → payment → warranty claims → follow-up. 863 commits in active daily production use.
Ground-up rebuild of FieldOps as a multi-tenant platform. Full Turborepo monorepo with tRPC end-to-end type safety, Drizzle ORM, and a drag-and-drop widget workspace. The architecture means the next service business deploys via configuration — not a new codebase.
Monitors 248 government URLs across all 50 states + DC, detecting content changes and feeding a RAG-powered chat interface. Built for a 1M+ customer business that needs accurate, up-to-date regulatory data without manual monitoring.
Automated policy reporting platform. The system discovers ads via a transparency API, a human approves targets via CLI gate, then Playwright workers navigate the real browser UI to submit reports — one job per unique advertiser. Ships with a Chrome extension for persistent session auth and residential proxy integration.
A community platform with two distinct technical achievements: a 1,041-line autonomous code modification engine that writes, reviews, and deploys its own updates via dual-LLM consensus; and a full community app with forums, messaging, push notifications, Slack/WhatsApp webhooks, RBAC, and a Claude Sonnet AI coach.
A fully automated daily intelligence briefing aggregating 13 data sources — calendar, email triage with AI-suggested actions, work email synced via Chrome extension (no OAuth needed), weather, headlines, and more — into one concise morning email. Every email action fires directly from the briefing via HMAC-signed links.
A custom Model Context Protocol server giving an LLM direct read/write/execute access to the entire infrastructure stack — deployments, databases, DNS, secrets, filesystem, GitHub, monitoring, and workflow automation. v4.1 routes routine queries to local Ollama at $0/query with cloud fallback.
A dedicated load-testing platform for the FieldOps ERP. Runs named scenarios against configurable endpoints with safe-mode by design — every write scenario creates tagged test records and cleans up on completion. Production data is never touched. Generates plain-English Markdown reports after each run.
A multi-model agent framework and testing harness. Routes tasks to the best available model — Claude, DeepSeek, local Ollama — with structured JSONL event logging, integrity checking, and a full test runner with factory state reset between suites.
31 purpose-built standalone tools for daily operations — planning, fitness tracking, scheduling, finance dashboards. No framework, no npm install, no build pipeline. Idea → deployed tool in under 2 hours average. The constraint is intentional: it proves you don't need a framework to ship useful software fast.
A fully self-hosted network infrastructure layer built around a dedicated bonding VPS. Four WireGuard tunnels bond all available WANs simultaneously from a travel router — ethernet, cellular, WiFi uplink, and tether. A private Tailscale mesh connects all servers, devices, and the router into a single zero-trust network accessible from anywhere. All servers are mounted as local drives on the dev machine.
Identify real operational pain, design the schema and API contract, build with AI pair-programming, deploy to self-hosted VPS, and iterate in production daily. Every system in this portfolio is running right now, handling real data for real businesses.
Zero SaaS overhead for core ops. Full control over data, infrastructure, and cost. When a platform owns your stack, they own your margins. Building it yourself means you own it completely — and you understand every layer when it breaks at 2am.
Not vibe-coding — structured. Every session has context via Master MD files, architecture docs, and a custom MCP server that gives the AI direct access to the codebase and infrastructure. The AI is a force multiplier, not a magic button.
Pushing further into AI-assisted development, self-hosted infrastructure, and systems that do real work. Side projects that scratch real itches — built for the love of shipping things that actually run.
Selectively accepting client work for custom AI brains and self-hosted business platforms — see /mind for the engagement we run most often. Happy to talk shop about systems architecture, AI tooling, or self-hosted infra.