Shawn Burrell — Production AI Operations in Practice

Opening

Reference material — not a résumé

Thirty years of B2B sales leadership. Prior to contracting the eight-week initial development phase for MaineCoastLife.com, I completed an intensive six-month AI Development and Business Integrations mentorship with Guy Tasaka — LMA Innovator of the Year, Technology Columnist for Editor & Publisher Magazine, 35+ years spanning The New York Times, Calkins Media, GateHouse Media, and WideOrbit.

I used the leadership skills I've acquired across my career to adapt into a new vertical, which has given me a field of expertise in both AI Development and Sales & Operations. Many have one or the other. Very few have both. The platform is now live and generating revenue under a signed annual licensing agreement with service-level commitments. This page is a short tour of what runs under the hood and the architectural pattern I use to run AI as infrastructure rather than as autocomplete.

About the licensee

An enterprise SaaS platform licensed to Peter Newton

Former Vice President at GateHouse Media (one of the largest local-newspaper groups in the United States) and former Chairman of the Board at evvnt (events-platform technology used by thousands of local publishers). The counterparty on this agreement is a sophisticated media operator who evaluates platforms for a living.

01 · The Platform

MaineCoastLife.com — by the numbers

An enterprise SaaS platform in production. Postgres + serverless edge functions + row-level security. A fleet of 30 venue scrapers runs nightly, feeding a two-stage pipeline into the events database, which drives the public site, the newsletter infrastructure, and the overnight audit layer.

Venues scraped

nightly, full fleet

3,175

Events in database

662 upcoming today

98%

Image coverage

3,114 of 3,175 events

3,736

Active subscribers

newsletter + alerts

46,071

Newsletter sends

lifetime

5,142

Scraper runs

lifetime, auto-logged

Live production numbers pulled from the MaineCoastLife.com Postgres database at build time of this document (Oct 31, 2026).

Ingestion pipeline — Supabase (Postgres + Deno edge functions), Resend for email

02 · Multi-Surface AI Operations Model

Three surfaces, one architecture

I split AI work across three distinct surfaces. Each has its own role, context, and authority. The specific tools are swappable; the architecture is tool-agnostic — portable to any enterprise AI stack.

Each surface has its own cost, latency, and risk profile. Routing is the architecture.

Why this matters: keeping these roles distinct is how AI stays useful instead of becoming either a toy or a liability. Most teams try to do everything with one AI surface and inherit all three risk profiles at once.

03 · Architectural Safety Boundaries

The differentiator

In one line Most teams treat prompt engineering as usage. I treat it as architecture — the safety posture lives in a system that enforces it even when nobody's watching.

The execution surface in my stack cannot read coordination-layer context — a documented fact about how these tools interact. Without a safeguard, an execution agent can produce confident, well-written changes that violate contracts the coordination layer knows about but the execution layer doesn't.

The gate turns a documentation convention into an enforced architectural boundary.

The pattern emerged from a real production incident and now prevents an entire class of latent-violation failures. Alongside the gate, I maintain a cross-referenced library of operational rules (~70 and growing) so the pattern scales from one operator to a team.

04 · Admin Console

Custom-built platform surfaces

Every management surface was designed and built by me. These are live screenshots of the production MaineCoastLife.com admin console as it runs today:

MaineCoastLife admin dashboard — platform overview with subscribers, events, revenue, and growth chart

01 · Operations Dashboard

The platform at a glance

Unified operations surface — active subscribers, upcoming events, business partners, and monthly revenue in one view. Quick actions for the most frequent admin tasks (add event, approve business, send newsletter, view analytics) are one click away. Subscriber growth trends over the last 30 days drive editorial and acquisition decisions.

MaineCoastLife admin — Events management screen with filters and status

02 · Events Management

646 upcoming events, one view

Unified events surface across all 30 venues with filtering by status (Pending Submissions / Upcoming / Past / Incomplete Scrapes / Drafts), category, and venue. Per-event ticket, view, and featured status. Every event here is the result of the automated ingestion pipeline — no manual data entry needed.

MaineCoastLife admin — Site Analytics dashboard with top events and businesses

03 · Site Analytics

89,995 page views to date

Custom analytics surface measuring platform performance: total page views, ticketed events, stories published, active subscribers. Top-events and top-businesses charts drive editorial and advertising decisions. All measured against the platform's own data — not a third-party analytics bolt-on.

05 · Security Posture

Security built in, not bolted on

The platform ships with enterprise-grade security controls as design primitives, not afterthoughts. The licensee's background — former VP at GateHouse Media, former Chairman at evvnt — meant the agreement had meaningful security expectations from day one.

Database

Row-Level Security

Every table enforces RLS policies at the Postgres level. Access control lives in the database, not the application layer — compromised application code cannot read what the database won't return.

Authentication

PKCE OAuth flow

Supabase Auth with Proof Key for Code Exchange. Self-registration disabled; all admin accounts manually provisioned by the platform owner. Protected routes gate every /admin/* path.

Transport

Automatic TLS

Netlify-managed Let's Encrypt certificates, automatic renewal, HSTS enforced. No manual certificate management, no expiry risk, no certificate-storage attack surface.

Edge

CDN & WAF

Netlify's global CDN with built-in DDoS protection and an OWASP-aligned Web Application Firewall. Static assets served from edge locations; dynamic calls go through scoped edge functions.

Data Integrity

Idempotent ingest

The scraper fleet operates against a hash-based dedup contract — the same input never produces duplicate database writes. Normalization enforces canonical shape; enrichment never bypasses it.

Backup & Recovery

Automated off-site

Supabase automated daily backups with 30-day retention, stored off-site. Point-in-time recovery available. The platform's SLA to the licensee includes contractual recovery commitments.

IP Control

Owner-controlled source

All source code, AI pipeline logic, database schema, and edge functions remain the sole property of Next Revolution Media, LLC under the licensing agreement. No third-party contributors to the production codebase.

Observability

Audit trail by default

Every scraper run, every pipeline stage, every newsletter send writes to persistent audit tables. 5,142 scraper runs on file. 46,071 newsletter sends recorded. Nothing happens off the record.

06 · Lifelong Learning — Receipts

Mentored, not self-taught

Prior to contracting the eight-week initial development phase for MaineCoastLife.com, I completed an intensive six-month AI Development and Business Integrations mentorship with Guy Tasaka — Managing Partner at Tasaka Digital, Technology Columnist for Editor & Publisher Magazine, LMA Innovator of the Year for creating the first FAST channels at local TV stations and newspapers. His 35+ years span The New York Times, Ziff-Davis, Calkins Media, GateHouse Media, and WideOrbit.

The mentorship was structured as weekly one-hour sessions — all recorded — with assigned tasks to complete before the next meeting. Learning the tools required to build and launch real projects was non-negotiable; it was a core part of Guy's training methodology, not an optional add-on. By the time I contracted the initial development phase for MaineCoastLife.com, the strategic frame and the hands-on capability were both in place.

What I learned to use and ship with during and after the mentorship:

Postgres, relational schema design, row-level security
Serverless edge functions (Deno/TypeScript), API design, authentication flow
HTML / CSS / JavaScript, Python, Git workflow
Email infrastructure, cron scheduling, autonomous agent orchestration

Currently shipping: a Business Owner Dashboard product; venue-health automation improvements; backup access automation; Linear ticketing stand-up for structured development workflow.

The rarer accomplishment. As a solopreneur, fewer than 1% ever take an enterprise concept all the way to market and revenue. I've done exactly that — and it follows a pattern that has run through my whole career: find ways to innovate and generate products, workflows, and solutions that help companies grow. The MaineCoastLife.com platform is the current chapter, not the opening one.

The point isn't the stack list — it's the combination. Sales & Operations expertise with real production AI development capability. Many people have one or the other. Very few have both. Lifelong learning isn't a tagline; it's an operating pattern.