The same approach, twenty years apart
Technology changes.
The method for success does not.
These two case studies, spanning 20 years—from a Java-based enterprise system for semiconductor manufacturing in 2005 to a Python AI pipeline for legal case files in 2026—use different technologies but demonstrate the same underlying approach: a proven methodology that delivers lasting results.
The hidden cost of premature solutions
Everyone knows process matters. So why am I talking about this so much?
There is always a point in a project when the process starts to feel like a luxury—or worse, unnecessary. People want to do the right thing; that is rarely the issue. The challenge is that everyone—including me—likes solutions. It is human nature.
Once a few possible solutions begin to emerge from a partially understood problem, it is easy for a team to become convinced it knows enough to start designing. Momentum builds, others naturally follow, and it feels like progress. But the hidden cost can be enormous: months of work, significant investment, and teams moving fast in a partially wrong direction—only to discover much later that a critical constraint was missed.
This is what I bring to a project: the ability to move quickly without losing the discipline that produces results. People bring experience, creativity, and ideas to a project, and the best solutions often emerge along the way. I make the process transparent, give teams room to explore and take the occasional tangent, while keeping the work anchored to the business requirements and constraints. The goal is not to make people follow a process—it is to create enough structure for the project to run smoothly while giving the team genuine ownership of the solution.
Side by side
Two projects, one method.
2005
Run-to-run control for low-volume products in a wafer fab
Seagate Technology, wafer fabrication. Process control system architect, Six Sigma project lead.
Most of the fab's products ran too rarely for automated process control to learn from them, so they were reworked until they passed. A Six Sigma project and a small change inside the control system I had built let them borrow the high-volume product's data. Twenty years on, it's still running.
- Six Sigma
- DMAIC
- Process control
- Java
2026
A local, air-gapped AI pipeline for legal document discovery
A county government legal team. Principal consultant, Improving.
Hundreds of pages of scanned records, read line by line. A month later: indexed, searchable and diffed case files that never leave one MacBook, with every fact traceable to its page.
- Discovery
- AI with guardrails
- Python
- Local models
1Start with the people
2005 A mind map and a system map with the four photo engineers before any design.
2026 Hallway interviews with the legal team before any tool search.
2Find the improvements that matter to the business
2005 An objective agreed with photo engineering: low-volume overlay within 10 percent of the high-volume product, with rework and tool capacity as the secondary measures.
2026 Four outputs the legal team named as immediately useful: every provider, a sourced timeline, problems and medications, and a keyword index.
3Identify the constraints as early as possible
2005 The high-volume product must not be touched. The engineers must own the groups.
2026 Nothing leaves the building. Nothing unverifiable goes to court.
4Make targeted changes toward the ultimate goal
2005 One deterministic change inside a control system the team already ran.
2026 Small deterministic steps in code, with the model doing one narrow job.
5Validate the solution as you go, reviewing with the business
2005 A gage study before trusting the data, ten new tests with all 179 existing tests still passing, then a month on live product with the engineers watching the results.
2026 Tests on all the Python, the 60-page golden document the team assembled run every night, and every output checked against its source page.
6Include the teams in every step for full ownership
2005 Alerts, a safe fallback, and training for the engineers.
2026 Provenance in every file, and a skill so a two-person team can extend it.
Other work
The rest of the resume
Full write-ups take time. These are the other projects behind the resume, with the stack up front for anyone scanning for Kafka, Spring Boot or React. I can discuss these at length as well, if interested.
2024 – 2026
County government application modernization
Sole on-site consultant on an application that had received only security patches since 2018. Upgraded Java 8 to 25, Spring Boot 2 to 3 and Hibernate 5 to 6, and rewrote the Angular 10 frontend in React 19, with zero outages or rollbacks. Built the team's local development environment and test coverage, and wrote the AI-assisted development guidelines they check their work against.
- Java 25
- Spring Boot 3
- React 19
- Docker
- Cypress
- JUnit
- Claude
2020 – 2024
Event-driven order management platform, Best Buy
Lead architect for the greenfield migration of a legacy order management system to Kafka Streams and Avro. Set the repository structure, core schemas and standards, then built the platform libraries for encryption, retries, error handling, state management and schema evolution that about thirty developers across domain teams build and operate their services on. Ran weekly Kafka forums and built a React and Spring Boot search tool for Kafka data still used by more than 200 people.
- Apache Kafka
- Kafka Streams
- Avro
- Spring Boot
- React
- Kubernetes
2018 – 2020
Customer-facing BestBuy.com applications
Led the team building profile management, automotive accessory scheduling and Total Tech Support registration, in React and Redux over an Express backend-for-frontend and Spring Boot services.
- React
- Redux
- Node.js
- Spring Boot
2012 – 2018
Consulting engagements, Object Partners
Employee-facing tools for previewing, requesting and scheduling site changes on BestBuy.com; a content management system that published product pages as near-real-time JSON; legacy data exposed as REST microservices on a pair-programming team; a customer loyalty rewards system for a restaurant chain; and testing software plus an accessibility video for an assessment company's legacy application.
- Java
- REST
- JavaScript
- SQL
1998 – 2012
Process control system architect, Seagate
Proposed and designed the run-to-run control system that shares real-time measurement data across cleanroom equipment, deployed in four countries and still in use. Co-authored SEMI standard E-133.1 for semiconductor manufacturing, with an article in IEEE Spectrum.
- Java
- EJB
- Oracle
- XML
- SEMI E-133.1
Get in touch
What's on your mind?
A few sentences is enough to start a useful conversation. Just say what your thinking and we'll take it from there.