// SQL Server DBA  |  AI Systems Builder  |  Rio Rancho, NM

Richard McCrea Database. Evolved.

25+ years turning enterprise data chaos into engineered order. From Intel's manufacturing pipeline to credit union AI — I don't just manage databases. I build the systems that make them intelligent. Currently working with SQL Server 2025.

See if I'm your hire View my work
25+ Years Enterprise DBA
$100M Saved at MGM Resorts
200+ Databases Migrated
185K Members in Production AI System
5,000+ AI Queries Self-Learned & Growing
// Featured Work

What I've Built

At a financial institution with 185,000+ members I designed and built a five-database enterprise AI intelligence platform from the ground up. One person. No vendor. No template. Built the right way — iteratively, methodically, with full audit coverage from day one.

The system answers plain English questions about member data with documented sources, confidence scores, and a complete audit trail on every single interaction. It gets smarter every day it runs. And when a better AI model comes along — we drop the engine and put in the new one. The intelligence stays. It always stays.
5 Database Architecture
185K+ Member Records
5,000+ AI Queries Self-Learned
0.2% Mail Return Rate
100% Audit Coverage
$300+ Saved Per 10K Mail Pieces
// Component 01

The Keystone Bridge

Real-time PolyBase connection to the production core banking system. Configurable sync frequency per table. Self-healing queue for sync failures and schema drift. New fields detected automatically. Deleted fields archived with last known values. The source of truth for all current member data — always live, always fresh.

// Component 02

The Semantic Intelligence Layer

The Soul of the system. Every business rule, every semantic relationship, every learned query pattern, every domain vocabulary term lives here — permanently, independent of any AI model. Native SQL Server 2025 vector embeddings power semantic similarity search. Drop the LLM. Put in a better one tomorrow. The Soul does not move.

// Component 03

The Data Warehouse

The memory layer. Facts the source system never calculated and never will. Running totals, historical deltas, behavioral patterns, derived insight scores. Consecutive on-time payments. Lifetime interest paid. Rolling balance averages. Product acquisition timelines. Every sync cycle makes it richer. Keystone knows today. The warehouse knows everything before today.

// Component 04

The Security Gateway

Security is a join, not a door. Every query includes a mandatory security join enforced at the data layer. Unauthorized users get nothing — no error, no message, just nothing. Immutable audit log. Every interaction logged permanently with full lineage. Employee behavioral anomaly detection. Regulator-ready from day one.

// Component 05

The Walled Garden LLM

On-premises inference engine. One network connection — to the internal network only. No internet. No telemetry. No path out. Not a temporary limitation — an intentional architectural decision that gives a security story any regulator can approve. Every prompt filtered by the security gateway before it arrives. Every response logged before it leaves.

// Component 06

The Governance Application

C# WPF control center that manages the entire platform. PolyBase table registration. Sync frequency management. Schema change review. DEV-to-PROD promotion pipeline. Tolerance monitoring dashboard. Broken thread queue. User and role administration. Audit log viewer. No schema change touches production without passing through here first.

How a Query Flows Through the System

01 → User submits a plain English question. Windows username captured automatically.
02 → Security check. If the user is not authorized, the query stops here. Access event logged regardless.
03 → Semantic layer parses intent. Vector similarity search resolves terms against the domain vocabulary and entity graph.
04 → Query plan assembled. Pre-warmed execution path used if known. New path built, cached, and added to the learning graph if not.
05 → Data retrieved. Current values live from the core banking system. Historical facts from the warehouse. Security join applied to every result set.
06 → For complex responses — curated, permission-filtered context passed to the walled garden LLM. Response synthesized.
07 → Answer delivered with source citations on facts, confidence scores on inferences, full audit log entry written.
08 → Learning loop. Query pattern, execution time, and result quality fed back into the semantic layer. System gets smarter. Every time.
We Know, Or We Say Nothing
No guesses. No hallucinations. No approximations. If the system does not know the answer it says so — and tells you exactly why.
The Soul Lives in the System — Not the Model
Every business rule, every learned pattern lives permanently in the architecture. Drop the LLM. Put in a better one. The intelligence stays.
Every Answer Carries Its Reasoning
Facts cite their source. Inferences carry a confidence score. A regulator must be able to read the audit trail and understand exactly what happened.
The System Knows When It Is Wrong
Self-healing queues, not silent failures. When something drifts the system flags it, logs it, and asks for help. Broken threads are learning opportunities.
SQL Server 2025 Native Vector Search PolyBase C# WPF LLM Integration Five Database Architecture OLLAMA RAG Semantic Layer Audit Logging tSQLt NUnit AWS Migration PowerShell
// Career

25 Years of War Stories

Sep 2024 — Present
SQL Server DBA (Full-Time)
Sunward FCU — Albuquerque, NM
Designing and building a five-database enterprise AI intelligence platform on SQL Server 2025 with native vector search, PolyBase real-time sync, semantic learning layer, walled garden LLM, and full audit coverage. Leading AWS migration including S3 backup configuration. Authored SQL Server installation standards, implemented gMSA across all service accounts. Hired full-time based on results delivered as a contractor.
Jan 2024 — Sep 2024
SQL Server DBA (Contractor)
Sandia Laboratory Federal Credit Union / Robert Half Tech — Albuquerque, NM
Engaged as contractor to modernize SQL Server infrastructure and design a production AI query system from the ground up. Delivered results that led directly to a full-time offer upon company rebrand.
Sep 2019 — Mar 2023
SQL Server DBA
Clark County Water Reclamation District / Robert Half Tech — Las Vegas, NV
Managed 24 SQL Servers across DEV, TEST, and PROD. Migrated SQL 2012/2014 to 2016–2019. Set up PowerBI, deployed Reporting Services 2017, implemented DevOps automation across infrastructure. Optimized developer queries and reinforced security standards.
Aug 2013 — Sep 2014
SQL Server DBA
MGM Resorts International — Las Vegas, NV
Reduced backup costs by $100M by eliminating redundant offline storage through strategic redesign with Xerox. Managed SQL Server clusters (2000–2014). Built internal C# and PowerShell tools for diagnostics and performance tuning.
May 2011 — Nov 2012
ETL / BI Developer
Intel Corporation / Mindteck Inc. — Hillsboro, OR
Designed Intel's first SSIS package spanning the entire manufacturing pipeline for near real-time visibility. Built ODS staging across MySQL, Oracle, and SQL Server. Achieved sub-10-second report rendering with MicroStrategy teams.
Apr 1999 — Oct 2010
Systems Analyst / Project Manager / Developer
3E Company — San Diego, CA
11 years delivering mission-critical systems. Led DevOps and Agile adoption. Migrated 200+ Crystal Reports to SSRS. Built real-time productivity tracking, MS Dynamics integrations, and full SQL Server infrastructure.
// Design Philosophy

How I Think About Systems

The hardest question in system design is not "how do I build this?"
It is "am I building the right thing — and am I building it right?"
Everything else follows from answering that honestly.
// Principle 01

The Soul Lives in the System — Not the Engine

Every AI system I design stores its intelligence — every business rule, every semantic relationship, every learned pattern — permanently in the architecture itself. The AI model underneath is the heartbeat. It keeps things running. But it is not the life of the system. Drop it. Recreate it. Swap it for something better tomorrow. The Soul does not move.

// Principle 02

We Know, or We Say Nothing

Financial-grade systems do not guess. They do not approximate. They do not hallucinate a confident answer when the data is ambiguous. If the system does not know the answer, it says so — and tells you exactly why. That is not a limitation. That is the feature. Trust is built one honest answer at a time.

// Principle 03

Security is a Join, Not a Door

Access control enforced at the application layer can be bypassed. Access control enforced at the data layer cannot. Every query I design includes a mandatory security join. Unauthorized users get nothing. No error. No message. Just nothing. The permission check is built into the query itself — not bolted on afterward.

// Principle 04

Never Throw Away a Measurement

Storage is cheap. Missing history when a question gets asked two years from now is catastrophic. Every system I build accumulates facts the source system never calculated — running totals, historical deltas, behavioral patterns, trend lines. Keystone knows what is true right now. Only the warehouse knows what was true last year.

// Principle 05

The System Knows When It Is Wrong

Silent failures are the most dangerous failures. Every architecture I design includes self-healing queues, tolerance monitoring, and broken thread detection. When something drifts — a sync breaks, a field changes, a query returns unexpected results — the system flags it, logs it, and asks for help. Broken threads are not failures. They are learning opportunities.

// Principle 06

Every Answer Carries Its Reasoning

Facts cite their source. Inferences carry a confidence score. Every interaction is logged with full lineage from question to answer. A regulator must be able to read the audit trail and understand exactly what happened and why. If you cannot report on it, measure it, and audit it — it is not done.

// Principle 07

Build for the Domain You Have. Design for the Ones You Cannot Imagine.

The first domain is never the last. Every architectural decision is made with the understanding that other domains will follow. Loose coupling. Plug-in modules. No domain is an island. The system you build today should welcome the requirements of two years from now without requiring a rewrite.

// Principle 08

Keep It Simple So It Works

Complexity is the enemy of reliability. The most elegant solution is the one that does exactly what it needs to do and nothing more. Every extra moving part is a future failure point. The assembly line that runs cleanly every time is worth more than the impressive architecture that requires a specialist to keep alive.

// Capabilities

The Stack

Database
  • SQL Server 2000–2025
  • T-SQL / Stored Procs
  • HADR / Clustering
  • PostgreSQL
  • MySQL / Oracle
  • Performance Tuning
ETL / BI
  • SSIS
  • SSRS
  • PowerBI
  • MicroStrategy
  • Data Modeling
  • ER Studio
AI / Modern
  • LLM Integration
  • Natural Language → SQL
  • Vector Embeddings
  • RAG Systems
  • AWS / S3
  • Walled Garden LLM Design
Dev / Ops
  • PowerShell
  • C# / .NET / WPF
  • Python
  • Git / GitHub
  • DevOps / Agile
  • gMSA / Security
// Interactive

Is Richard Your Hire?

Let's Find Out in 2 Minutes

Tell me about your company, your data problems, and what you need. An AI trained on 25 years of Richard's actual work will tell you exactly how well he fits — and why.

Start the Qualifier
// Contact

Let's Talk

Phone
702-931-8424
Location
Rio Rancho, New Mexico
// Send a message — I'll get back to you directly
✓   Message received — Richard will be in touch shortly.