Matthew Colborn OS — real estate operating system
Architect & sole engineer · AWS · production
A single operating system that runs an entire real estate practice: lead capture and CRM,
automated nurture, transaction and escrow tracking, a client portal, an off-market marketplace,
public booking, and a project-management layer that turns goals into dated, sequenced work.
- Serverless API — Python Lambda services behind API Gateway, split by trust boundary:
a Cognito-authenticated admin API and a separate unauthenticated public API with its own least-privilege IAM role.
- Data layer — DynamoDB single-table-per-domain design across leads, contacts, calendar,
consent, suppression, tasks, and project records, with pagination-safe scans and idempotent writes.
- Scheduling engine — protected time blocks, a public booking flow that can't double-book,
and a read-only personal-life overlay mirrored into the shared calendar so automation never books over it.
Publishes a live ICS feed consumed by Apple Calendar.
- Project management — a deterministic SMART decomposer (no LLM dependency) that breaks a goal
plus a deadline into phase-templated tasks, backward-scheduled from the target date with dependency chains;
moving a deadline cascades every transitive dependent.
- Front end — a single-file dashboard on S3 + CloudFront: day/week/month views, an execution stat board,
domain filtering, and optimistic check-off with local-cache hydration for instant paint.
PythonAWS LambdaAPI Gateway
DynamoDBCognitoS3 + CloudFront
SES / SNSEventBridgeIAM
JavaScript
Compliance-grade messaging infrastructure
Architect & sole engineer · AWS · production
Outbound email and SMS in a regulated space is mostly a compliance problem, not a sending problem.
I built the consent and suppression layer that makes automated outreach defensible under TCPA and CAN-SPAM.
- Consent as a gate, not a flag — an append-only audit store recording the exact disclosure text,
version, timestamp, IP and user agent at the moment of opt-in. SMS additionally requires written-consent
proof before a single message can be sent.
- Three independent layers — per-contact consent, per-address suppression, and provider-level
suppression as a delivery backstop. A send has to clear all three.
- Bounce & complaint pipeline — SES notifications → SNS → a Lambda consumer that distinguishes
permanent from transient failures, auto-suppresses and revokes consent, with CloudWatch alarms on
bounce and complaint rates.
- Sensitive-data guard — outbound SMS is screened for SSN and payment-card patterns
(regex plus Luhn validation plus keyword context) to satisfy carrier content rules.
- Carrier registration — A2P 10DLC / toll-free registration driven through the API, and a
public opt-in page whose disclosure block doubles as the carrier-required proof of consent.
Amazon SESSNSLambda
DynamoDBCloudWatchTCPA / CAN-SPAM
A2P 10DLCSPF / DKIM / DMARC
PorchLight OS — brokerage operating system
Architect & lead engineer · Google Cloud · in development
A multi-user operating system for a real estate brokerage, built to compete with the incumbent
platforms on the workflows that actually drive production — accountability, pipeline, and agent activity.
- Node service tier on Google Cloud with Firebase Authentication and role-scoped access per agent and admin.
- Designed around integration seams to transaction management, MLS data and mail, so the system
reflects real activity rather than asking agents to re-enter it.
- Built in a separate cloud project and repository from my own systems, with the IP and licensing
boundaries defined up front.
Node.jsGoogle CloudFirebase Auth
Service accountsMulti-tenant
Branshe OS — trades business platform
Architect & engineer · Microsoft Azure · in development
An operating system for a cleaning-trades business, built on the Microsoft stack because that's where
the business already lived: marketing site → scheduling → invoicing → owner dashboard.
- Microsoft Graph and M365 integration for mail and calendar, so scheduling writes into the tools staff already use.
- Accounting integration for invoicing, and paid-acquisition instrumentation feeding the same pipeline view.
AzureMicrosoft GraphM365
QuickBooks API
Turn and Earn — isolated lead-capture tier
Architect & engineer · Azure · in development
A lead-capture product for a development group, deliberately built as its own billing, repository and
cloud tenancy — connected to my primary AWS platform through a single, versioned API seam rather than
a shared database.
- Astro front end deployed to Azure Container Apps, provisioned with Terraform.
- One integration contract between clouds, so either side can be handed off or sold without untangling the other.
AstroTerraformAzure Container Apps
Cross-cloud integration
Acquisition instrumentation
Engineer · production
Server-side and client-side conversion tracking across Meta, TikTok and Google, mapped to real
pipeline events rather than page views — so paid spend is measured against booked appointments and
closed business instead of form fills.
Conversions APIEvent mapping
AttributionAnalytics