The resume answers what I've done. This page answers what I'm doing it for. In a software-solved world a few years out, the work that matters most won't be writing software — it will be deploying production systems that put humans, animals, food, shelter, and community back at the center. This is what I'm building toward.
Kibble → real food. Dogs first, because that's where the system is rigged hardest and the cost of being wrong is most legible. Then humans, because there's no honest reason it stops at dogs. Bespoke meal kits generated from your animal's vet record, your family's history, your allergens, your seasonality — recipes that come out of a conversation, not a catalogue.
Sourced within a 60-mile radius. Not because hyper-local is fashionable — because it's the only supply topology that survives the bandwidth, cost, and energy economics of the next twenty years. The bottleneck isn't "who can grow good food." It's "who can route the right food to the right household reliably on a Tuesday in February." That's a software substrate problem, and it's a trust topology, and it's exactly the shape of what I built in Universal Interface.
Delivered milkman-style. The right cadence is recurring, relationship-grounded, low-friction. Not a marketplace. Not a SaaS. A route — a person, or later a small autonomous vehicle, that knows you, knows your dog, knows your garden, and shows up.
Agrihoods. Neighborhoods that grow some of what they eat and supply some of what their neighbors eat. Children in the garden their whole lives — proverbially, figuratively, and literally. The substrate that connects the meal kit to the soil is the same substrate that connects the household to the neighbor; that's not coincidence, that's the architecture.
Bespoke houses. An AI conversation that lands as a full CAD model plus BOM, supply list, structural validation, electrical plan, and permitting trace. Any house, any size, any quirk — off-grid cabin or apartment retrofit. Built by robots where robots make sense, and by humans against generated visual work instructions where humans do. Bespoke at the price of standard. The Alexander-Hamilton house for everybody.
Then planes, cars, boats, coffee, anything. The pattern is the same: voice in, hardware out, supply local, validation rigorous. The CAD-to-BOM-to-build pipeline that works for a house works for a fishing skiff, an electric ultralight, a roastery, a lathe. Once the substrate is real, every category of physical thing is a config.
Training systems for every servant-served dynamic. CopApp is one cell. The pattern — Dialog → World → Psyche — generalizes cleanly to doctor-patient, engineer-customer, teacher-student, parent-child, contractor-homeowner, lender-borrower. Anywhere there's a power-imbalanced authority relationship, the served party benefits from a rapidly-customizable simulation of the served-by party. That's a platform, not a vertical.
Mars is the limit case. A self-sufficient settlement on Mars is the same problem as a self-sufficient agrihood in Alabama, with a tighter constraint set and worse comms latency. If you can do the agrihood, you can do Mars; if you can't do the agrihood, "going to Mars" is a marketing exercise.
The human is the core for empowering. The healing is the heart. The proverbial garden is the goal.
Each project on the resume is the prototype for one piece of the bigger thing. They look like a portfolio because that's the medium I'm working in. They're not a portfolio. They are a roadmap.
Wedge for: bespoke meal kits, bespoke houses, bespoke anything. Any two systems become routable through a connector the AI writes, tests, and refines itself. The household ↔ farmer, household ↔ vet, household ↔ permitting office, household ↔ neighbor are all the same problem shape. Air-gapped routing by data sensitivity means it works for households that don't want their vet record and their dog's diet leaking onto somebody's marketing list.
Wedge for: voice-first, hardware-aware household substrate. One human, one dog, voice in, organized everything-else out. The agrihood version is one human, N neighbors, N suppliers, and the soil — same architecture, same offline-first SQLite mirror, same outbox / watermark reconciliation when connectivity is patchy. The "court-admissible narrative" angle in PawTrek becomes "neighbor-of-record narrative" in the agrihood.
Wedge for: training systems for every authority dynamic. The Dialog → World → Psyche pipeline is not police-specific. It's the shape every rapidly-customizable simulation of a power-imbalanced relationship needs. Doctor, teacher, engineer, parent. Get it right once, instantiate it everywhere. CopApp is the demonstration; the platform is what's behind it.
Wedge for: multi-agent simulation of any complex physical production. A meal-kit cohort needs Vet, Supplier, Route, Kitchen, Allergen, Family as console operators. A house build needs Architect, Supplier, Code-Inspector, Robot, Neighbor, Owner. The Flight Director pattern (a single human can preempt, hold, scrub, go) is exactly what a homeowner orchestrating a custom build needs.
Wedge for: safety-critical hardware-aware simulation, end to end. The same rate-hierarchy, plant-inversion, and Monte Carlo discipline that closes a guidance loop closes a structural margin loop on a generated-house design. The book is the doctrine that lets anyone — including a future hire — pick up the next system from the same vocabulary.
Wedge for: AI conversation → CAD + BOM + printability / machinability / moldability. The name is honest about what it is. Hydrodynamic calculations for hull shapes is the demonstration; the platform underneath is the bespoke-anything substrate. This is the seed of the "Alexander-Hamilton house for everybody."
Wedge for: hardware that has to actually be safe to occupy. NASA human-rating discipline, DOORS hazard tracing, V&V campaigns, IPT coordination across primes and international facilities, CFRP production for flight-rated hardware. The bespoke-house and bespoke-meal-kit world will need exactly this discipline when it touches real households. Most consumer-AI companies don't have it. I do.
Two honest callouts so I don't sound naive about this. Neither of them is the part most software people would name.
The AI-generates-a-meal-kit-from-your-dog's-vet-record part is solvable now. The "your neighbor can actually fulfill it reliably on a Tuesday in February" part is what nobody has cracked. The hard layer is trust and incentive, not algorithm. Universal Interface is the right shape for it because the system pair is household ↔ farmer, not farmer ↔ farmer. My HRO research on communication patterns in high-reliability organizations is the unfair-advantage layer underneath.
AI-to-CAD-BOM is the easy half. The hard half is the same V&V and hazard-trace discipline I ran in DOORS for human-rated spaceflight, applied to municipal code, structural inspectors, and electrical authorities. Robots can build it; nothing built without inspection sign-off is occupiable. That's the gap, and that's where I have unfair advantage.
The human is the core for empowering.
The healing is the heart.
The proverbial garden is the goal.