Work

Case studies and working systems across enterprise data architecture, team training, workflow automation, and applied AI.

Case Studies

DATA ENGINEERING

Enterprise Pricing & Revenue Infrastructure

Context

General Mills manages a $20B+ customer business and $5B in annual trade promotions across dozens of retail brands.

Problem

Pricing and trade promotion data lived across fragmented pipelines, creating latency and inconsistency during critical price advances.

Key Decision

Separate transactional pipeline execution from the analytical foundation. Anchor the business on a single governed data layer rather than disconnected reporting tables.

What Was Built

Production data pipelines using Spark and Airflow on BigQuery, serving over 30 business dashboards and analytics tools used across virtually all corporate price advances.

Result

Directly contributed to $60M+ in profit generation, supported $5B in annual trade optimization, and accelerated pricing decision workflows by 4x.

ENTERPRISE ARCHITECTURE

Enterprise AI Data Readiness Standards

Context

Across General Mills, more than 100 data engineers build and maintain data pipelines feeding analytical and operational systems.

Problem

Teams built pipelines to varying local conventions, creating inconsistent contracts and quality gaps that blocked reliable AI deployment.

Key Decision

Establish clear architecture standards at the foundation rather than attempting to clean and govern disparate data assets after models fail.

What Was Built

Comprehensive architecture standards covering data contracts and data modeling principles, paired with quality checks and the strategic rationale for shifting toward AI readiness.

Result

Adopted as the governing baseline across 100+ data engineers to ensure production data is structured for AI applications by design.

ENTERPRISE EDUCATION

Enterprise Training Program

Context

Technical professionals at different experience levels did not have a shared path for learning AI and data engineering.

Problem

Adoption efforts kept stalling because there was no common foundation to build on.

Key Decision

The curriculum needed to build judgment, not just introduce tools.

What Was Built

Built and ran a two-year internal curriculum for 50+ engineers and scientists at General Mills.

Result

Participants established a shared foundation and practical standards for applying AI and data engineering in production.

SYSTEM DESIGN

LOI Automator

Context

A real estate investor spent five hours every week drafting repeat letters of intent.

Problem

The legal structure and boilerplate stayed the same, but property details and deal terms changed with every offer.

Key Decision

The investment decision needed to remain human. The repetitive document creation did not.

What Was Built

A system that accepts deal parameters in plain language and generates a finished, properly formatted letter of intent.

Result

A completed letter of intent in under two minutes, saving five hours each week.

Personal Systems

Smaller builds for my own life that show the same method: find the real problem, pick the simplest tool that fits, ship something that gets used.

LOCAL AI INFRASTRUCTURE

Vectorized Knowledge Base & MCP Server

Problem

Years of journals, technical notes, and strategic decisions lived in unstructured files, making historical recall slow.

What Was Built

An embedding pipeline and vector store connected to a custom Model Context Protocol server, allowing local AI agents to query personal history in plain English.

Result

Instant semantic retrieval across years of decisions and journals with zero context lost.

WEB APPLICATION

Verse & Journey

Problem

Consistent daily study requires historical context and structured reflection without getting lost in multiple commentaries.

What Was Built

A focused web application that automates the daily study workflow and reflection prompts.

Result

A consistent daily habit with zero prep friction.

verse-and-journey.lovable.app→

RAPID PROTOTYPE

Responsibility Through Play

Problem

A $500 calendar system was a real commitment. I wanted to know if turning responsibility into a game would actually work for my son before buying it.

What Was Built

A simple task tracker built in a weekend, with emoji icons, daily resets, and animations when he completes a task.

Result

My son used it on his own and took ownership of it. I validated the idea before spending anything.

Related story: Building Responsibility Into Play

DIGITAL PRODUCT

Premarital Counseling Experience

Problem

Counseling curriculum was scattered across physical binders and loose documents, making sequential guidance difficult.

What Was Built

A guided web application structured around the exact sequence and pacing of the counseling curriculum.

Result

A focused digital experience that eliminated manual material prep and kept the guidance personal.

Have a System That Needs Building?

Whether you need enterprise team training, advisory on data foundations, or a keynote speaker for your event, let's talk.

Start a Conversation

← Back to the homepage