Selected Work

These case studies explain why projects mattered — not just what was built. Each one is organized around a problem, an approach, and lessons that generalize beyond the specific context.

The work spans thirty years: from the first Linux server on a Pennsylvania university campus in 1993 to platform strategy in regulated medical device development today. The technologies changed. The underlying questions — how do platforms create leverage, how do organizations learn, how does technical work become lasting capability — did not.

Infrastructure & Open Systems

Cutter: Democratizing Linux on Campus — 1993–1996

Before Linux was institutionally acceptable, a single machine running on commodity hardware gave nearly 2,000 campus users their first experience with open systems — and became the campus's first web server. The story of how friction, not merit, determines technology adoption.

Platform Is Relationships: Voyager, KLN, and the Consortium That Grew Beyond Itself — 1994–1999

A library consortium built tools for migrating legacy systems to a new ILS. Those tools spread across Pennsylvania, to Georgia, and eventually into every Voyager deployment nationwide — not because of marketing, but because the consortium had relationships that carried the work. On why platform thinking requires human infrastructure, not just technical architecture.

Data, Analytics, and Enterprise Systems

Enterprise Registration & Data Warehouse — 1999–2002

The university had enrollment data locked in a mainframe. Building an analytical layer alongside it — without replacing it — gave Academic Affairs tools for evidence-based admissions and retention planning at a time when comparable institutions were still working from static reports. On why information only becomes strategic when it is connected.

Teaching, Research, and Institution Building

Building Engineering Programs & ABET — 2002–2022

Starting from a university with no engineering infrastructure, five degree programs were created — including the first ABET-accreditable computer engineering program in Pennsylvania's state university system. The programs were built before most of the approvals existed. On building infrastructure before the mandate arrives.

Teaching to Last: Technology-Agnostic Engineering Education — 2002–present

Twenty years of teaching embedded systems as the hardware changed every five years. The embedded systems textbook, the Fabrication Lab, eleven new courses: all built around the same principle — teach the engineering methodology, not the current chip. On why what you teach should outlast the hardware you teach it on.

Constraint Generation and the Commitment to Evidence — 2002–2008

A Ph.D. completed over six years while teaching a full load, commuting to UMBC twice weekly, and personally funding the degree. The research: making OWL reasoners honest about what they actually know, grounded in a corpus of 200,000 real ontologies. On intellectual honesty as an engineering discipline.

Discovering What the Algorithm Knows: Interpretable ML and the Limits of Opacity — 2003–2005

Research into evolutionary discovery of composite SVM kernels — producing models that were both more accurate and human-readable. Published at AAAI-05 (17% acceptance rate). On why AI tools that practitioners can understand and reason about are more valuable than black boxes that merely score well on benchmarks.

Platform Strategy & Regulated AI

Platform Strategy in Regulated Environments — 2022–present

In regulated software domains — medical devices, aerospace, defense — the organization that generates trust continuously will outperform the one that generates evidence episodically. On why governance and velocity are the same problem stated differently, and why platform investment compounds over time in ways that first-product metrics cannot capture.

AI Transformation in Regulated Development — 2022–present

Generative AI moved the bottleneck from code generation to verification — inside IEC 62304, FDA design controls, and 510(k) submission. On adopting AI without exempting it from evidence: continuous governance, interpretability as a design requirement, and the same intellectual honesty that shaped earlier ML and knowledge-representation work.

The recurring thread: systems that outlast their builders, tools that spread because they solve real problems well, and organizations that become more capable than they were. The doctrine that connects them.