Nick C. AI Automation Engineer
Available for new projects Hire on Upwork

n8n · Claude API · Make.com · GoHighLevel

Automation that still works when nobody's watching it.

I build the unglamorous parts: the pipeline that sources and verifies leads overnight, the agent that answers the call, the classifier that decides what happens next. Systems that run unattended and tell you when they don't.

Rate
From $50/hr
Engagement
Project or retainer
After delivery
30 days monitoring, free
Contracted via
Upwork

Services

What I build

Most of this work looks the same underneath: something has to happen reliably, on a schedule or on an event, without a person in the loop. The domain changes; the engineering doesn't.

Selected work

Four systems, in production

Each of these runs live. I can walk through execution history, retries, and real records landing — not a screenshot of a workflow canvas.

  • Lead generation

    Automated outbound pipeline

    • ProblemManual prospecting burned hours a day and still sent to dead addresses.
    • BuiltSourcing and bulk enrichment, email verification gating, global suppression, per-lead scheduling, and a circuit breaker that halts sending when delivery signals go bad.
    • ResultThe list builds and sends itself; the operator reviews replies instead of spreadsheets.
  • Voice AI

    Real-time voice agent

    • ProblemPractice and qualification calls needed a counterpart that responds in real time, not a script.
    • BuiltLive audio pipeline — speech in, model reasoning, speech out — with turn handling tight enough to feel like a conversation, and full transcripts stored per session.
    • ResultA running system anyone can call and be pushed back on, with the transcript waiting afterward.
  • Applied AI

    Reply classification and routing

    • ProblemInbound replies arrived faster than anyone could triage, and the wrong auto-response is worse than none.
    • BuiltA classifier that assigns intent and disposition, drafts the response, and holds it for approval instead of sending blind.
    • ResultTriage happens automatically; a human still signs off on anything that leaves the building.
  • Payments

    Usage-based billing and referrals

    • ProblemPer-module access and metered usage don't fit a single subscription record.
    • BuiltEntitlements granted per module, Stripe webhooks made idempotent so redelivery can't double-charge, grace windows on failed payment, and referral attribution on top.
    • ResultBilling state survives outages and retries, and can be rebuilt from Stripe if a webhook is ever missed.

These were built for my own companies, which is why I can show you the inside of them. On client work I'll show whatever the contract allows, and say plainly when it allows nothing.

Process

How a project runs

  1. A call

    Twenty minutes to hear what breaks today and what it costs you. If automation isn't the answer, I'll say so on the call.

  2. Scoping

    A short questionnaire covering access, tools, and edge cases, plus a walkthrough of exactly what you'll get and what I need from you.

  3. Build

    I build it, test it against real data, and document what I touched so you're not locked to me.

  4. Handoff

    You get a working system, a walkthrough, and the documentation. Then thirty days of monitoring at no charge while it settles.

  5. Ongoing, if you want it

    After thirty days, monitoring and changes move to a monthly retainer. Plenty of clients don't need one — that's fine.

Stack

What I work in

n8nMake.comGoHighLevelClaude API OpenAI APILiveKitDeepgramSupabase PostgresNext.jsStripeRailway Apollo.ioInstantlyResendGoogle Workspace

Next step

Tell me what's eating your week.

If it happens more than once and follows rules, it can probably run without you. Send me the messy version — I'd rather see the real process than a tidied-up description of it.