Founding and scaling an AI department

A defense engineering and R&D firm wanted applied AI but had no team, standards, or delivery record in it. The capability had to win and deliver DoD programs while producing reusable intellectual property, so the company was not simply selling engineering hours one contract at a time.

• Founded the department as its first engineer and owned AI/ML strategy from problem definition through production deployment.
• Hired and grew the team to ten core engineers; led programs of 30+ across engineering, operations, business development, contract SMEs, and government stakeholders, directing the project managers who ran them day to day.
• Worked directly with ORBIS's C-suite on AI strategy, funding, and headcount, and represented the work to the Navy, shipyard stakeholders, partners, and prospective customers through proposals, briefings, and sales pitches.
• Set technical standards, prioritized the research portfolio across concurrent DoD programs, and presented programs to customers, executives, and audiences of up to 200.

1. 1

   Tie headcount to demonstrated value

   Each hire expanded a delivery engine that customer-funded programs had already proven. The department accumulated its mandate through contracts, reusable IP, and reduced costs, which kept company leadership's trust and gave every new engineer meaningful work on arrival. Hiring ahead of funded work was the option rejected.

2. 2

   Protect roughly a third of capacity for research

   Engineers needed room to test architectures and build reusable components beyond a single contract. I documented the path from research experiments to funded programs so the allocation could survive budget reviews; that research produced the copyrighted IP that later became the main source of R&D revenue.

3. 3

   Hire for range across the system

   Applied AI crosses data acquisition, evaluation, software integration, security, and domain validation. Engineers with depth plus range could follow a problem across layers and catch downstream consequences early, which avoided a department built around handoffs.

4. 4

   Formalize standards before the team outgrew shared context

   As the team approached ten engineers and programs passed thirty people, evaluation practices, deployment requirements, security controls, and architectural decisions were documented so they no longer depended on the conversations where they started.

• A servant-leadership operating model: I owned priorities, resources, customer alignment, and organizational friction; engineers owned their areas and were expected to explain assumptions, test conclusions, and surface risk early.
• Leading leaders as well as engineers: the project managers on each program worked under my direction, with clear decision rights, cover with the customer, and early notice when priorities changed.
• Engineers were brought into the full customer problem, including the existing workflow, what failure would cost, and how success would be measured.
• Failures were reviewed by examining the data, requirements, architecture, and decision path that produced them, which kept people willing to report uncertainty early.
• Team members taught one another, which spread expertise and reduced dependence on any single person, including me.

• AI department scaled from 1 to 10 core engineers; programs of 30+ led across departments, contractors, and government partners.
• Additional multimillion-dollar DoD programs in segmentation, anomaly detection, predictive analytics, and maritime sensor analysis.
• Proprietary, copyrighted AI/ML IP supplied roughly 80% of ORBIS R&D revenue.
• Three consecutive DoD contracts, approximately $7M, procured and led.
• 5/5 performance rating for five consecutive years.

I build AI organizations the way I build systems: decide where the leverage is, earn scope with evidence, and write down the standards before the team needs them. Expect protected research time, headcount requests backed by delivered value, and engineers who own real decisions.

• Article: From a team of one to a defense AI program
• Resume (PDF)

Confidentiality: Program names, customers, and contract details beyond the public descriptions above are withheld under DoD and company confidentiality obligations.