Boardroom to bench — the combination is the offering. I build evidence-centered systems at the seam of defense industrial, critical-infrastructure, and resilient-operations work.
Translator-operator is the shortest name for it: someone who sits between the factory floor, the data model, the fielded system, and the executive room — and keeps the boundaries honest in every direction. I lead programs where strategy, funding, technical substance, and operational reality have to meet in the same head.
Each lane combines strategic framing, program execution, and operational fluency. The strongest fit is wherever those three need to meet at national-program scope. Specific role types listed under each lane — not exhaustive; close adjacencies welcome.
Where to bet, how to bet, and how to keep the AI layer honest. Source-grounded retrieval, knowledge architecture, AI assurance — and the discipline that keeps LLM enthusiasm from becoming LLM theater.
Defense-relevant manufacturing and strategic materials. Mineral-to-system supply-chain tracing, ownership graphs, jurisdictional risk, industrial capacity. National-program scope.
Manufacturing modernization, supply-chain resilience, OT cybersecurity, value engineering, physical AI on the factory floor — the lane EY @ MxD already operates in, extended to higher-scope program leadership.
Programs that are too large, too multidisciplinary, and too cross-stakeholder for one technical function to own. Translate strategy to execution and back, board to bench.
Mission-relevant capture, growth, and offering development that needs a leader fluent in both the technical substance and the operating model. Build the lane, fund the lane, execute the lane.
The portfolio isn’t a list of products to ship. It’s a deliberate convergence: every plane — the executive seat, the independent lab, the graduate lineage, and the long-term aim — points at the same kind of work.
Senior Manager at EY (Industrials · Advanced Manufacturing). Program Manager for the EY Digital Operations Hub @ MxD, where the hub floor includes three Yaskawa and two Universal Robots (UR) industrial arms, physical-AI cells, and the working bench for client workshops in manufacturing modernization, supply-chain resilience, OT cybersecurity, workforce, and value engineering. MxD — Manufacturing × Digital — is the national Digital Manufacturing Institute, a Manufacturing USA institute sponsored by the U.S. Department of Defense, where industrial modernization and defense-industrial resilience are worked in the open. National in scope, executive in seat, operational in feel.
Instruments, embedded platforms, sensor work, knowledge-graph experiments, and active robotics builds. This is not a portfolio of products to ship; it is continuous, credible literacy in the dialect engineers actually use. The bench exists so the boardroom conversation stays grounded — and so I can sit at an engineer’s workbench and talk about deliverables, not just dashboards. The bench’s first public artifact is Sprout (MIT) — a plant-monitoring platform that refuses to report a number it didn’t measure.
The M.S. in Computer Science — earned at Northwestern’s Institute for the Learning Sciences — started with LISP on a state-of-the-art robot rover — two sensors: a black-and-white 240×160 camera and a bump sensor. Roger Schank’s questions — memory, scripts, plans, cases, expectations, understanding — are still the right questions. C, Smalltalk, and the languages between then and now sit on the same arc. Today’s LLM / RAG / graph systems are the modern shape of the same problem.
Evidence-centered systems for defense-relevant manufacturing, critical materials, and resilient operations — at national-program scale, with provenance preserved end-to-end and human judgment kept in the loop. The work, the lab, the lineage, and the aim aren’t four hobbies; they’re four faces of one deliberate convergence.
I would rather have a small thing that works and is honest about what it doesn’t know than a large thing that performs confidence. These six axioms are the spine of the work.
Every claim needs a traceable source. Data is not a number; it is a number from a specific source at a specific time with a specific confidence. The chain applies to sensor readings, supply-chain assertions, and AI outputs equally.
The interesting scenarios are degraded links, failed sensors, incomplete data, adversarial environments. Offline-first behavior, dual-mode messaging, and graceful degradation are primary requirements — not edge cases.
AI assists discovery, extraction, and reasoning. It does not replace provenance. AI-generated content is labeled and never promoted to authoritative without source validation. The boundary is preserved on purpose.
Logging supports reconstruction of what happened, not only real-time diagnosis. If you cannot recreate the event exactly, the event is unrecorded. Raw data is preserved on principle.
Whether the action is hardware scan approval on an instrument or a decision-grade output in a supply-chain workbench, the system supports human judgment rather than replacing it. Automation is transparent and reversible.
A system that cannot be rebuilt is only partly understood. Playbooks, install notes, restore points, driver evidence, known-good commands are not paperwork after the fact — they are how the system proves it can survive interruption.
Most stacks treat physical sensing, communications, AI, applications, and integration as separate concerns. The point of the work is to converge them — while preserving the chain of evidence at every step.
The graduate-school lineage is not decorative. The questions early AI cared about — memory, representation, planning, cases, expectations, understanding — are still the questions worth caring about now.
Understanding is not text similarity. Understanding requires source-grounded structure: roles, goals, plans, procedures, cases, constraints, confidence, and evidence.
The master’s was earned entirely inside the Institute for the Learning Sciences, studying under Roger Schank — and it started concretely: a state-of-the-art robot rover with two sensors — a black-and-white 240×160 camera and a bump sensor — programmed in LISP. Schank’s questions arrived at the same time as the hardware did.
Smalltalk and C followed; later languages followed those. The point was never the syntax. The point was learning to read the systems underneath — how representation, control, and memory actually behave when the machine is running.
LLMs make extraction easier. They do not replace conceptual structure. Vector recall + graph enrichment + source grounding is the modern shape of the same problem the rover and the ILS reading lists were already chasing.
The arc lands at programs that connect sensors, provenance, critical materials, industrial modernization, and decision intelligence — with the AI layer kept honest about what it is and isn’t.
A public-facing subset. A complete record — including sensitive authorizations and field-safety credentials — is available through verified recruiting channels.
Strongest fit: defense-industrial base programs, mission-first industrial-tech, critical materials and supply-chain resilience, manufacturing modernization at national scale, and roles that need a translator-operator at the seam.
hello@vkhogue.comEvidence path — selected examples, sanitized work samples, and a full résumé are available through verified recruiting or professional channels.
Complete credential and authorization details are available through verified recruiting channels. The public site is intentionally restrained on sensitive material.