Open to remote roles/Chicago/Senior engineer and tech lead
Patrick Finnigan
I build AI agentsand the tools around them.
Eight years shipping web products, most recently at Beacons AI, where I helped grow Beam, the creator AI agent, about 12× in one summer and built the Beacons MCP from scratch, so Claude and ChatGPT can work with Beacons directly. TypeScript by default.
BuiltLink-in-bio editing tools for the agent, and one parameterized analytics tool in place of many per-metric ones.
LearnedShowing the agent a rendered view of the page, not just its style settings, made its edits much better.
ResultAbout 1,000 to 12,495 weekly active creators over summer 2026.
AI agentstool designevals
12,495weekly active creators, Oct 2026
~1,000Start of summer
12,495Oct 2026
Automation
clautopilot
Headless Claude Code automation that takes a ticket to a pull request, then keeps answering review comments until it's ready to merge. Works across GitHub, GitLab and Linear.
Claude CodeGitHub / GitLab APIs
Agent setup as code
dotclaude
My Claude Code setup as code: hooks, skills and a git-synced memory shared by several coding agents across two machines, with evals that check the agents actually use it.
hooksskillsevals
Experience
Where I've worked.
2022 – 2026
Beacons AI remote
Senior Software Engineer and Tech LeadApr 2025 – Oct 2026
Helped grow Beam, the creator AI agent, about 12× in weekly active creators. Built the Beacons MCP from scratch and shared its tools with Beam to keep the two at parity. Tech lead for Beacons for Brands: architecture, tickets, code review, and delegating to engineers and AI agents.
Software EngineerJun 2022 – Apr 2025
Built the React and TypeScript frontend and Python APIs for branded email, which grew adoption 750%. Ran onboarding and checkout A/B tests with LaunchDarkly and Amplitude.
2018 – 2022
Riviera Partners San Francisco
Engineering ManagerSep 2021 – May 2022
Managed three frontend engineers on the core recruiting app, reporting to the CTO.
Engineering Team LeadJan 2020 – Aug 2021
Led the team and architected its work, including a real-time sourcing interface that points recruiters to unreviewed candidates.
Software EngineerAug 2018 – Dec 2019
React and Ruby across the recruiting app, including a client-facing KPI dashboard.
Every AI assistant I use shares one memory of me, and it runs on an old laptop in my apartment. Claude on web and phone, Claude Code, Codex, ChatGPT and Grok all read and write the same long-term memory. A small MCP server (MCP is the standard way AI apps call outside tools) runs on a 2018 MacBook Pro, reachable only through a Cloudflare Tunnel behind an authenticated Worker gateway. I deploy it by asking for a deploy in chat.