chris-alarcon open to FDE · Solutions · Implementation — Remote US
proof of work, not a resume

I build the system first, then make it run without me.

AI automation engineer with 8+ years shipping enterprise workflow systems — and a public fleet of production agent systems I designed, built, and operate myself: content pipelines, job-search engines, outreach machines, and ops bots running on Claude Code, n8n, and Python, every day, unattended.

40+production integrations shipped
800+runbooks & technical docs authored
8+yrsenterprise systems delivery
24/7agent fleet in production today
01

Systems in production

real pipelines I run today — not demos, not tutorials
multi-agent operations

The Agent Fleet

A dedicated Mac mini running a fleet of Claude-powered agents on cron lanes — content research, job scouting, cold outreach, inbox triage, and a Telegram ops bot I command remotely. Agents write to git, sync a shared knowledge vault, and report results without supervision. Designed with the boring stuff that makes agents survivable: quality gates, dedupe stores, structured-output schemas, audit logs, and self-verification steps.

Claude Codesub-agentsMCP cron / launchdTelegram botgit-synced memory
cron ──▶ agent lane ├─ scrape + verify ├─ quality gate └─▶ notion / telegram
5+ lanesrunning daily, unattended
data pipeline · llm scoring

Job-Search Engine

Multi-source ingestion (LinkedIn via n8n, Google Jobs API, RSS, Apify actors) → SQLite dedupe with a rolling window → hard-filter rules → LLM scoring against my actual resumes → QA agent verifies live URLs → delivery to Notion. Full JSONL audit trail so a "zero results day" is diagnosable in ten seconds. The same discovery → filter → verify → deliver shape I'd ship for a customer.

Pythonn8nREST APIs SQLiteLLM-as-judgeNotion API
4 sources ──▶ dedupe ├─ hard filters ├─ llm score vs resume └─▶ apply-ready queue
100%auditable — every drop logged
content operations

Media Company of One

A gated, multi-phase content pipeline that runs like software: trend radar scrapes TikTok/Instagram/YouTube, an idea well ranks topics, then phase-gated skills (hunt → blueprint → write) with contract files between phases so nothing skips a checkpoint. Scripts, thumbnails, descriptions, scheduling — agent-assisted end to end, with my voice and judgment as the only manual steps.

agent orchestrationphase gatesscraping APIs evalsprompt systems
radar ──▶ topic lock ├─ blueprint gate ├─ script + assets └─▶ scheduled publish
3 gatesno phase runs without its contract
end-user automation

Tenant Radar

Rental-lead automation for a real property: an email receptionist agent that reads inbound leads and drafts warm, screened replies with a booking link into Gmail Drafts (human approves every send), plus a scout that watches local Facebook groups for qualified seekers. Hits flow to Notion and Telegram. Small system, real stakes, zero missed leads.

Gmail APIscrapinghuman-in-the-loop NotionTelegram
inbound lead ──▶ classify ├─ 5-question screen ├─ draft reply + link └─▶ human approves
2 cronsreceptionist + scout, live
02

The enterprise half

why regulated industries trust me in the room

Eight and a half years inside a Tier-1 bank running the exact loop a forward-deployed engineer runs with customers: discovery → POC → integration → rollout → enablement — across Risk, Compliance, and Engineering stakeholders who don't forgive sloppy systems.

  • Led end-to-end implementation of enterprise workflow systems across 5+ business units — requirements to rollout to training.
  • Built 40+ production automations with n8n, Power Automate, and Alteryx — REST APIs, webhooks, OAuth, SQL — with retry branches, error handling, and failed-run recovery.
  • Authored an 800+ document library of runbooks, SOPs, API configuration notes, and resolution guides used by multiple teams during launches and incidents.
  • Cut manual reporting 60%+ and operational inefficiencies 30% by replacing human glue with systems.
03

Working stack

used in production, not from a course

AI / Agents

Claude Code & agent SDK patterns · sub-agents · MCP servers · structured outputs · evals & quality gates · prompt systems · LLM-as-judge scoring

Automation

n8n · Power Automate · Alteryx · cron/launchd orchestration · human-in-the-loop approval flows · failure recovery design

Integration

REST APIs · webhooks · OAuth · JSON · SQL · SQLite · Gmail/Notion/Linear/Telegram APIs · Apify & scraping infrastructure

Code

Python pipelines · JavaScript/Node · git-based multi-machine workflows · shell · Astro/static sites · Cloudflare Pages

Enterprise

Regulated-environment delivery · stakeholder discovery · POC → rollout · documentation systems · training & enablement

Communication

Public technical teaching on YouTube · runbooks & SOPs · docs that unblock the moment a user gets stuck

04

Built in public

watch me explain the systems I ship

YouTube — Chris Builds Systems

Weekly walkthroughs of real AI automation systems: agent architectures, n8n workflows, Claude Code builds, and the practical operator notes nobody writes down. If you want to know how I think and communicate before a single call — it's all on camera.

Watch the channel

christopheralarcon.com

Written work hub — guides and breakdowns on turning one-off AI prompts into repeatable workflows technical users can actually operate.

Read the site

Hiring for forward-deployed, solutions, or implementation work?

I'm the person you put in front of the customer who can build the integration, fix the workflow, write the runbook, and teach their team — in the same week. Remote US (NJ, ET).