Automation Infrastructure Portfolio

Engineering Philosophy

Case Study: BrewBot v1 — From Idea to Working Prototype

The challenge:


Businesses needed a simple way to automate lead generation and onboarding without expensive software or complicated tooling.

What I did:

  • Studied multi-agent AI workflows

  • Mapped the user journey from first touch to booked call

  • Built a working prototype powered by Espresso Engine™ on Relevance AI

  • Designed the UI and brand

  • Integrated it with a live website and lead capture flow

Result:

  • A functioning AI assistant that captures leads, answers questions, and automates early-stage workflow tasks—showing how an “AI employee” can plug into real-world businesses.

  • Prototype used to secure a partnership with Property Navigator

Erik Warren

Automation Infrastructure Engineer

Production-grade automation systems designed, deployed, stabilized, and maintained in live operational environments.

Focused on deterministic workflow architecture, reliability engineering, and scalable AI-integrated systems.

Primary Stack
Make.com · REST APIs · Webhooks · OAuth · JSON · Relevance AI · Google Workspace

Core Principles
Deterministic branching · Failure isolation · Structured outputs · Observability · Scalable workflow design

I approach automation as infrastructure, not as a collection of disconnected workflows.

Every system I build is designed with:

  • Deterministic routing logic

  • Explicit state control

  • Failure-path isolation

  • Structured data contracts

  • Observability through logging and traceability

Automation should be predictable under load, debuggable under failure, and extensible without introducing cascade risk.

My focus is long-term maintainability and production stability, not rapid one-off builds.

Core Systems Overview

SystemEnvironmentScopePrimary FocusBrewBot Ultra – AI Lead Review EngineProductionAI-driven intake + routingDeterministic AI orchestrationZiprent Automation StabilizationEnterprise Production1,200+ accountsFailure reduction + infrastructure rebuildKC Prime – Internal AI Ops FrameworkOngoing DevelopmentModular AI ops systemStructured AI workflow architecture

System 1: BrewBot Ultra – AI Lead Review Engine

Problem Context

Manual lead review created inconsistent response times, subjective decision-making, and missed high-intent opportunities.
The system required structured intake processing, deterministic routing, and controlled AI decision logic.

Architecture Overview

Event-driven intake pipeline with structured AI decision layer and deterministic routing control.

Infrastructure Stack

  • Make.com (primary orchestration layer)

  • Webhook-based intake

  • Google Sheets (structured datastore + logging layer)

  • Relevance AI (LLM decision engine)

  • SMTP integration (controlled outbound routing)

Workflow Architecture

  1. Webhook intake receives structured form submission

  2. Input normalization and field validation

  3. Deterministic status tagging

  4. AI decision engine call (structured JSON output required)

  5. Confidence-based routing (Respond_Now / Nurture / Manual_Review)

  6. Controlled outbound email execution

  7. Full logging of decision metadata and routing path

Reliability Mechanisms

  • Enforced structured JSON outputs from AI layer

  • Explicit routing tokens to prevent ambiguous branching

  • Status locking to prevent duplicate processing

  • Failure fallback path with manual review trigger

  • Scenario-level error handling modules

  • Logging of intent level, urgency, and confidence score

Scalability Design

  • Modular scenario separation

  • Expandable routing categories

  • Structured data schema allows API migration beyond Sheets

  • Designed for horizontal expansion into multi-channel intake

Ownership Scope

  • Full system architecture

  • Workflow design

  • Error-handling implementation

  • Ongoing iteration and optimization

  • Production monitoring