# Matt Perry — Full Site Corpus

> The complete content of https://www.mattperry.info/ in one markdown document, published for LLM and agent consumption. Last updated 2026-09-10. Contact: matt@mattperry.info · https://linkedin.com/in/matthew-perry-27606 · Raleigh, NC.

## Who

Matt Perry — engineering leader and hands-on builder. Navy veteran; ships code daily.

25 years across full-stack development, cloud architecture, and applied AI. Directs a ~50-person organization shipping production agentic-AI systems — and still ships production code daily, with an evidence-based view of where AI-assisted engineering helps and where it doesn't.

**At a glance:** 25+ years building · ~50-person org directed · 3 cruise-line brands running his agent platform in production · 23 years at SAS before that.

**Now:** Director of Technical Automation & Technical Integrations at Holland America Line (Carnival Corporation), remote from Raleigh, NC. Builds and operates a multi-tenant agent platform — Java/Spring Boot, React/TypeScript, PostgreSQL/pgvector, Terraform on AWS — where agents, knowledge bases, tools, and multi-step flows are versioned configuration rather than code.

## The AI on this site

The site hosts "Ask My AI" — a live agent grounded in Matt's resume and supporting career documents, running on the same kind of KB-grounded platform Matt builds professionally. Anyone (human or agent) can ask questions or paste a full job description for a candid fit assessment, gaps included. Answers cite their sources.

How it works, in five steps: a knowledge base (the documents, embedded as vectors and retrieved by meaning) is wrapped as a read-only search tool; an agent (Claude) plans and batches its own searches — every claim must trace to a retrieved passage, and gaps are stated plainly; a flow is the front door that runs asynchronously while the page polls; and every search, passage, reasoning step, token, and cost is traced and recorded. Nothing is a black box.

This Q&A box is deliberately a small use case of a much bigger platform Matt designed and built, which covers: multi-agent orchestration (parallel branches, routing, agents calling agents as tools, delegation governance); tools and integration (REST tools generated from OpenAPI specs, AWS Lambda, document extraction, MCP endpoints published and consumed); grounding (retrieval with citations per knowledge base, chunking and retrieval tuning, vision OCR for scanned documents); trust and governance (tenant isolation enforced in the database, role-based access, guardrails, PII redaction, schema-validated structured output); quality (LLM-judge eval suites with regression gates, human review loops, every run pinned to its exact configuration); and economics and operations (every model call metered for tokens and cost, prompt caching, pluggable model hosting, end-to-end OpenTelemetry tracing). Everything above is configuration, not code — the same runtime serves every use case, from this Q&A box to multi-agent production workloads across three cruise-line brands.

**API access:** POST https://www.mattperry.info/api/ask with JSON `{"question": "..."}` (optional `conversationId`, 8–64 chars `[A-Za-z0-9_-]`, for multi-turn context). Response: `{"answer": "...", "citations": ["..."]}`. Question cap 20,000 chars; 5 requests/minute per IP; answers typically arrive in 10–25 seconds.

## Experience

### Director — Technical Automation & Technical Integrations · Holland America Line (Carnival Corporation) · October 2024 – Present

- Directs ~50 professionals — managers, project managers, technical leads, onshore/offshore engineers — with multimillion-dollar capital budgets; owns hiring, performance reviews, and career development.
- Built and operates an internal agent platform (Java/Spring Boot, React 19/TypeScript, PostgreSQL/pgvector, Terraform on AWS ECS Fargate and RDS) where operators configure agents, knowledge bases, tools, and multi-step flows as versioned data — no code deploy needed to change behavior. In production across Holland America, Princess, and Seabourn.
- Owns the platform's entire cloud footprint: Terraform-provisioned ECS Fargate across two availability zones, RDS Multi-AZ, edge TLS and certificate lifecycle, secrets management, monitoring, disaster recovery, and cloud spend.
- Built an OpenAPI-to-agent-tool pipeline: published API specs are imported and turned into governed, callable tools for LLM consumers.
- Publishes and consumes Model Context Protocol (MCP) endpoints in production with per-tenant authentication.
- Leads the team's adoption of AI-assisted development — tool evaluation, usage patterns, guardrails for code review, data handling, and IP — and ships production code the same way.
- Designed the governance layer enterprise AI requires: per-tenant row-level isolation, RBAC, content-hashed config snapshots pinned to every run, input/output guardrails, and an eval harness with LLM-judge scoring and regression gates.
- Pioneered the enterprise AI strategy adopted across multiple Carnival brands and championed by executive leadership; established corporate-wide API standards and a Java microservice layer handling 10M+ daily transactions.

### Senior Manager — Guest Commerce · Holland America Line · April 2024 – October 2024

- Standardized ad-hoc automation into scalable multi-vendor processes.
- Delivered revenue-critical guest-facing systems (Shore Excursions, "Come Back and Sail") generating hundreds of millions in annual revenue.
- Built an AI-powered crew-onboarding proof of concept whose demonstrated business value led directly to promotion and the creation of the Technical Automation team.

### SAS Institute · Cary, NC · 2001 – 2024 (23 years)

**Principal / Advisory Pre-Sales Technical Architect (2016–2024):** Drove $10M+ in annual pipeline across 50+ strategic engagements with Fortune 500 clients (BNSF, McDonald's, FedEx, Carnival, Travelport, UPS). Built custom analytics web applications embedded in SAS Viya — replaced an incumbent Palantir solution at BNSF, delivered a price-optimization UI for McDonald's ($823K), and real-time bot detection for Travelport ($3.6M) scoring 3,400+ transactions per second. Created the Viya Data Editor (evaluated by SAS R&D and Dr. Goodnight) and SAS Console. Built a shore-excursion recommender deployed to 26 cruise ships. Technical Account Manager for Carnival Corporation.

**Principal Training Consultant (2011–2016):** Ranked #1 SAS instructor globally in 2014 and 2015. Authored the SAS Demand-Driven Forecasting curriculum; taught across the US, Helsinki, and Barcelona.

**Senior Software Developer & Consultant (2001–2011):** Led front-end development for SAS Drug Development and Contact Center Planning & Optimization; architected BI dashboards for healthcare, finance, and utilities clients.

### United States Navy · Sonar Watch Supervisor & Sonar Technician · 1993 – 1999

First sailor aboard to qualify as Sonar Supervisor during a first enlistment; led a team of 10 technicians. Deployed for Operation Sharp Guard (NATO enforcement of UN sanctions in the Adriatic). Qualified Enlisted Surface Warfare Specialist (ESWS); awarded the Navy and Marine Corps Achievement Medal and a Letter of Commendation.

## Skills

Front end & web (React, TypeScript, Vite, accessible UI, structured data and crawlability for AI agents) · APIs & identity (REST, OpenAPI/Swagger spec-driven tool generation, OAuth 2.0/OIDC, Okta, Microsoft Entra ID, scoped service tokens, RBAC) · AI engineering (Anthropic Claude, AWS Bedrock Converse, agentic orchestration, RAG chunking/retrieval evaluation, MCP publish and consume, LLM-judge evals with regression gates, guardrails, per-call cost attribution) · AI-assisted development (Claude Code in real delivery, team adoption patterns, guardrails, impact measurement) · Cloud & platform (AWS: ECS Fargate, RDS PostgreSQL, S3, Lambda, CloudFront, ALB, Secrets Manager, CloudWatch, ECR; Azure Entra ID; Terraform; Docker; CI/CD; multi-AZ HA; OpenTelemetry) · Languages & data (Java/Spring Boot, TypeScript/JavaScript, Python, SQL, PostgreSQL/pgvector, SAS) · Leadership (hiring, performance, career development, roadmap ownership, agile delivery, vendor/offshore management, budget, executive communication).

## Recognition

- **2024–25** — AI strategy championed by Carnival's executive leadership; recognized with the founding of the Technical Automation team and promotion to Director
- **2023** — Replaced a deeply entrenched Palantir solution at BNSF with the modern CIQ (Continuous Improvement in Quality) analytics application
- **2021** — Top SAS Customer Advisory Performer of the Year — the "Technical Decathlete" award
- **2019** — Patent contributor — outlier detection system
- **2018** — $3.6M Travelport win — real-time bot detection at 3,400+ TPS
- **2014–15** — Top Global SAS Instructor, two consecutive years
- **1993–99** — Navy & Marine Corps Achievement Medal and Letter of Commendation — U.S. Navy

## Education & Certifications

MBA, Information Technology Management — North Carolina State University · BBA, Computer Information Systems (Honors) — James Madison University · AWS Certified Cloud Practitioner · Advanced & Base SAS Programming

## Always building something

Nights and weekends Matt runs Collection Innovations (https://collectioninnovations.com) — a production SaaS for collectors (AI valuations, image recognition, 100k-item catalog) built solo on the same stack he leads at work: Spring Boot, React, Postgres, Terraform on AWS, and heavy AI-assisted development.

## Machine-readable surfaces

- https://www.mattperry.info/llms.txt — index for LLM consumers
- https://www.mattperry.info/resume.md — markdown resume
- https://www.mattperry.info/Matthew_Perry_Resume.pdf — canonical resume (PDF)
- https://www.mattperry.info/api/ask — grounded Q&A agent (POST, JSON)
- schema.org/Person JSON-LD embedded in the HTML page