Systems & chief engineer · AI agent teams & automation · Wichita, KS
I'm John Dempsey, a systems and chief engineer with 25+ years in aerospace and defense. I build multi-agent AI teams and the automations around them, engineer Microsoft Azure and Microsoft 365 systems, and set the AI architecture and strategy that hold it all together, with the requirements and traceability that serious systems need.
30 minutes with me. You leave with a clear next step for your agent team, automation or AI architecture.
Where AI programs stall
Experiments multiply, but nobody has picked the workflows where automation actually pays, or measured results in dollars and hours instead of demos.
Data, identity, integrations and failure modes are an afterthought, so the pilot can't scale into the environment you actually run.
Requirements don't trace to data, tests and results, and the first customer review, audit or certification milestone stops the program cold.
What I do
I've spent my career taking systems from requirements through integration, test and production. I bring the same discipline to AI: the agent teams and automations that do the work, the Microsoft platform they run on, and the architecture and strategy around them.
Requirements through integration, test and production, the way I've run aircraft, missile and sensor programs, applied to the AI and business systems you run today.
I build and run my own team of specialized AI agents every day, and I can build one around your operations and the automations you need.
I architected and administer a Microsoft cloud environment end to end. I can design, automate, harden or untangle yours.
How AI fits into the systems you already run, designed in from the start rather than bolted on.
Move from AI experiments to AI operations: pick the workflows where automation pays, then make adoption durable.
Written blueprints your team can build from and run without me in the room.
I'm the fixer before the auditor.
Readiness work comes with concrete fix recommendations, not just a pass/fail pre-assessment. I map your AI to the frameworks that apply (DO-178C, ARP4754B, ARP4761A, ISO 26262, ISO/PAS 8800, IEC 62304, NIST AI RMF, ISO/IEC 42001, the EU AI Act) and build the traceability and evidence reviewers ask for.
I don't issue certificates. Approvals come from the relevant authority, notified body or registrar.
Solution guides & insights
Practical guides on agent teams, Microsoft environments, AI architecture and regulated AI. Published guides live on AI Assurance Directory; more are on the way.
Where state AI legislation stands and what it means for teams deploying AI.
Read the article GuideA path through DO-178C, ARP4754B, ARP4761A and the emerging guidance for machine learning in airborne systems.
Read the guide GuideDoD Responsible AI, CMMC and NIST SP 800-171 for contractors adopting AI.
Read the guideRoles, routing, hand-offs and human supervision for a multi-agent team, drawn from the one I run.
Coming soonAzure, Entra ID and Microsoft 365 groundwork before you roll out AI to your organization.
Coming soonChoosing the workflows where automation pays and the governance that makes adoption stick.
Coming soonExample build: an agent-run directory
AI Assurance Directory is one example of what I build: a register of AI assurance tools mapped to the standards they answer to, researched, fact-checked and maintained day to day by my AI agent team with me in the loop. Every mapping links to an official source.
Directory counts as of 1 November 2026.
Take the free readiness self-checkPick your sector for a starting point.
Cross-industry frameworks most organisations using AI meet first.
A starting point, not a determination of applicability. Your scope, market and contracts decide what actually applies. EU AI Act high-risk obligations do not apply yet: Annex III systems from 2 December 2027 and Annex I systems from 2 August 2028 (Regulation (EU) 2026/1744).
Process
30 minutes. Your goals, your environment, your constraints and any deadline: a launch, an audit or a certification milestone.
Requirements, workflows, platforms and obligations. You see where AI pays, what it must connect to and what applies.
Strategy, architecture and a written blueprint, with risks ranked so effort goes where it matters.
I build it with your team, or hand over a blueprint your engineers can run.
Proven results
Strategy is only as good as the engineering behind it. From my work as a fractional CIO:
The Problem
A fast-growing learning-management software company ran on a 10-year-old database that couldn't sync with its mobile inventory app. Manual data entry and subscription costs were piling up every month.
The Solution
I modernized the database with Python middleware and a cloud backend, creating real-time bidirectional sync between the database and the mobile inventory app.
The Result
60% reduction in processing time and $2,000/month eliminated in subscription costs and manual data entry. The system now processes 10,000+ transactions a day with zero sync errors.
About
I'm John Dempsey, and I run Sparky's AI from Wichita, Kansas. I've spent 25+ years in aerospace and defense: project engineer on F-15 programs at Boeing, principal systems engineer at Raytheon, chief engineer at ATK, director of program management and now Deputy Chief Engineer on aviation defense programs. I also serve as fractional CIO for a learning-management software company, where I set the AI strategy and architected the Microsoft cloud environment I administer.
My discipline is systems engineering: DO-178C and DO-254 systems, the ARP4754B and ARP4761A development and safety processes, FAA Part 23 and DoD certification, where every requirement is traced and every claim needs evidence. I bring that to the team of AI agents I build and run every day, to automations and Microsoft systems, and to AI architecture and strategy.
I hold a BS and MS in aerospace engineering, an MS in mechanical engineering and an MBA. My agent team runs AI Assurance Directory, one of the systems I've built. Sparky's AI is where I build the same kind of systems, agents and automations for your organization.
john@sparkysai.comFAQ
A set of specialized AI agents, each with a defined role, coordinated by a router and supervised by a person. I run one myself: an operations router plus agents for research and fact-checking, coding, agent supervision, mail and travel. I design the roles, hand-offs and oversight so the team does useful work you can check.
System-level thinking. I start from requirements, interfaces and failure modes, then choose the AI, data and platform that fit. You get an architecture your team can build, test and run, not a demo.
Yes. I architected and administer a Microsoft cloud environment end to end (Azure, Entra ID, Exchange and Microsoft 365), including the automations that keep it running.
No. Certificates and approvals come from the relevant authority, notified body or accredited registrar. I help you prepare the plans, traceability and evidence they review.
AI Assurance Directory is one example of what I build: a cited register of AI assurance tools mapped to standards and frameworks, run day to day by my AI agent team with me in the loop. Sparky's AI is the consulting side: hands-on help with systems, agent teams, automations and architecture for your organization.
No. This is engineering and technology consulting. For legal interpretation of a regulation, work with your counsel.
Book a 30-minute call, or email john@sparkysai.com. If you'd like a head start, read the solution guides or take the free readiness self-check on AI Assurance Directory.
Book 30 minutes. Bring the system, the environment and the deadline, and we'll work out the architecture and the first steps.
Not ready to talk? Take the free AI readiness check.