I build the whole system, and it ships.
Applications, data, integration, infrastructure, security, identity: I design the architecture, write the code, and run it in production. Four platforms are live, one of them on a manufacturer's shop floor every shift.
Real production, not prototypes: paying customers on that shop floor daily, a multi-tenant platform spinning out its own products, a local-first AI stack running on GPUs I own, and 800+ passing tests across four platforms.
- 10 yrs
- shipping software
- 4
- production platforms
- 6
- industries entered cold, all shipped
- 800+
- tests across the portfolio
- In productionOH SCRAP!Multi-tenant SaaS · Manufacturing ops
- LiveSolved The ProblemPlatform engineering · Multi-tenant SaaS
- Operating dailySovereign Command CenterAI infrastructure · Full-stack platform
- In productionKAEB Sales SystemBusiness operations platform
4 platforms I architected, built, and operate
A decade across healthcare, manufacturing, marketing data, logistics & CAD
- Kaeb Sales
- Sutcliffe Mechanical
- Knight-Swift
- Autodesk Gold Partner
Bring in the operator, not just the coder.
Engagements are fixed-scope or fractional, and everything ships with tests, docs, and handoff runbooks. Here's what you get.
Own the whole stack
Schema to UI, and everything between: data model, tenant-safe RBAC, real-time sync, payments, cloud, and deploy. On OH SCRAP that meant estimating, scheduling, a live shop floor, and QuickBooks sync, end to end.
- Data model & RBAC
- Real-time & APIs
- Payments & integrations
- Cloud & deploy
Ship reliably, leave it hardened
Real paying customers in daily production, backed by 800+ passing tests across the portfolio, pre-commit gates, and observability. You get a green suite and a runbook, not a support ticket.
- 800+ tests passing
- Pre-commit gates
- Sentry / Pino
- OpenAPI docs
Ramp fast on your domain
EMT-turned-engineer: I learn the messy parts of a business first, then build software people actually use, across healthcare, manufacturing, marketing data, logistics, and CAD.
- Healthcare / EMS
- Manufacturing
- Marketing data
- CAD / Salesforce
See it work before you commit.
Start with a money-back MVP (a working proof of your idea in 40 hours, or your fee back), then scale into a full build. Fixed scope, your code in your accounts.
MVP Proof-of-Concept
Not ready to commit to a full build? I'll turn your core idea into a working proof-of-concept in 40 hours, and if it doesn't do what we agreed, you don't pay.
See how it worksPlus a $4,500 diagnostic, 45-day build sprints, project rescue, and fractional retainers: see all engagements.
Four production platforms, built and operated solo.
Three are my own products, one a confidential client build. Real software in daily production, not decks or prototypes. Each ships tested and documented (runbooks, not lock-in) and opens to a full case study.
…and a decade more
The competency map, with receipts.
Organized the way an architecture review is organized, not the way a résumé is. Every cluster names the systems that prove it, and every item on it is running in something today.
Application architecture & delivery
Schema to UI, in production, four times over. Typed React front ends on Node/Express services, plus a schema-driven CRUD framework that generates grids, forms, and API routes from declarative config, so a new module is configuration instead of bespoke code.
- React 18 / 19
- TypeScript
- Next.js
- Node / Express
- Schema-driven CRUD framework
- Turborepo monorepo
- Module entitlement (client + server)
Proof: KAEB · OH SCRAP · Solved The Problem · Sovereign
Data architecture & retrieval
Tenant isolation as a data-model decision, not a query filter: database-per-tenant on Solved, per-organization partitioning on OH SCRAP. Plus BigQuery warehouse work and a MongoDB-backed vector store that indexes code by symbol kind, line range, export status, and typed import/call edges.
- MongoDB / Mongoose
- Database-per-tenant
- Per-org partitioning
- BigQuery datasets & pipelines
- Embeddings + vector store
- AST-aware chunking (ts-morph)
- Graph retrieval (imports / calls)
- Package-level change-impact analysis
Proof: Solved The Problem · OH SCRAP · Sovereign Vault · Colling Media
Integration & middleware
The systems that have to talk to systems you do not control. Accounting sync, payment rails with idempotent webhooks and explicit rollback, real-time transport on the shop floor, CRM and CAD platform extension, and an MCP server that exposes an internal platform to outside agents under scoped permissions.
- QuickBooks Online sync
- Stripe Connect + webhook idempotency
- Payment state machines & rollback
- Socket.io real-time
- REST / OpenAPI contracts
- Salesforce admin & development
- Autodesk Fusion 360 plugins
- MCP server (tool exposure, audit events)
Proof: OH SCRAP · Solved The Problem · NexGen (Autodesk Gold Partner) · Sovereign
Infrastructure: self-hosted and cloud
Services run on hardware I own, not only on someone's managed tier. A two-tier local fleet (a workstation node with a 24GB RTX 3090 in an external enclosure, a laptop node with a 6GB A1000) serves models under Docker, alongside GCP, Render, and Vercel for the hosted platforms. Sizing and failure modes come from operating it, not from a vendor datasheet.
- On-premises / self-hosted deployment
- Docker / docker-compose
- Self-hosted GPU fleet (24GB + 6GB)
- Capacity & quantization sizing
- Google Cloud (Pro Cloud Developer, earned 2019)
- Cloud Storage + signed URLs
- Render blueprints · Vercel
- Sentry + Pino observability
Proof: Sovereign (own hardware) · OH SCRAP (GCP) · Solved (Render) · Colling Media
Security, governance & audit
Controls written on the assumption that the operator, or the model, will eventually get it wrong. Secrets are redacted at the model boundary against 12 vendor-anchored patterns before code leaves the machine, and over-redaction is the chosen failure mode. Data classification is treated as an architecture constraint here, a habit from handling protected health information on an ambulance rather than from a compliance certificate.
- Secret redaction at the model boundary
- Four-stage policy engine (regex / AST / vector)
- Single-use approval tokens, payload-hash bound
- Elevated tier for .env / Dockerfile / credential paths
- Hash-chained, tamper-evident audit events
- Docker execution sandbox
- Spend caps (daily / monthly, warn / pause / block)
Proof: Sovereign Command Center · OH SCRAP (tenant-safe RBAC)
Identity & access
Auth0 as the identity plane across four platforms: OIDC for people, machine-to-machine credentials for services, and a JWT to tenant-resolution to just-in-time-provisioning to module-entitlement chain on Solved. Authorization is enforced on every route, model, and service, never only in the UI.
- Auth0 OIDC
- Machine-to-machine (M2M) credentials
- JWT → tenant resolution → JIT provisioning
- Cross-tenant Auth0 federation
- Tenant-safe RBAC (route / model / service)
- Per-seat scopes & org resolution for agent tools
- Ed25519 org-bound licensing
Proof: Solved The Problem · OH SCRAP · KAEB · Sovereign (MCP seats)
AI platforms for private deployment
Local-first LLM orchestration on owned hardware, with escalation to a cloud model only behind an explicit override flag. Model serving, retrieval, orchestration, sandboxed execution, and cost control are separate replaceable components, which is what makes a stack assessable instead of a single vendor bet.
- Ollama model serving
- 7B / 8B / 14B local; 32B at roughly Q4
- Quantization trade-offs & VRAM budgeting
- AST GraphRAG retrieval
- MCP orchestration
- Anthropic SDK (explicit escalation)
- Sandboxed execution (Docker)
- Cost estimation & spend enforcement
Proof: Sovereign Command Center (7 subsystems, 88 tests)
Structured evaluation & quality engineering
Technology gets picked the way code gets tested: define the criteria, run the candidates on real hardware, score them, then write the recommendation with the trade-offs stated. A benchmark service scores 5-6 local models on my own fleet so the sizing claims are measured, and retrieval eval, index coverage, and drift monitoring keep the answer current after adoption.
- Model benchmarking & scoring service
- Retrieval eval & index coverage
- Drift monitoring
- Adversarial / red-team payload generation
- Build vs buy vs fix verdicts
- Vitest / Jest (800+ tests portfolio-wide)
- Husky pre-commit gates
- Swagger / OpenAPI
Proof: Sovereign eval service · KAEB (639 tests, zero failures) · OH SCRAP (90+) · Operations X-Ray
From an emergency department to Software Tailor.
It started with nurses keeping CPR time on a paper towel. The throughline ever since: walk into a messy operation and replace the paper with software people actually use.
Emergency Department· EMT → self-taught developer
Phoenix, AZRunning CPR, I watched nurses keep time on a paper towel. I went home, taught myself Java, and built an Android documentation app: CPR timer, med timer, IV access, airway. That was the start.
- Java
- Android
- Healthcare / EMS
NewTek· Web Developer
ArizonaBuilt and maintained WordPress sites, then realized web apps paid better and solved harder problems, and went looking for them.
- WordPress
- PHP
- Web
Independent / agency work· Application Developer
RemoteLaravel 5.x business tools: quoting engines, technician management, logistics systems, and financial / wealth tools like asset allocators and more.
- Laravel
- PHP
- Business tooling
Colling Media· Developer
ArizonaA digital + traditional marketing agency. Built and maintained LeadFlow, stood up BigQuery datasets, and shipped recruiting software and tools for Knight-Swift Transportation. Worked daily in GCP; earned the Google Cloud Professional Cloud Developer certification in 2019.
- GCP
- BigQuery
- Node
- Knight-Swift
Winamac Coil & Spring· Software Developer
IndianaA spring factory. Built shop-floor QA/QC tools and internal quoting systems: the manufacturing-ops DNA that later became OH SCRAP.
- Manufacturing
- QA / QC
- Quoting
NexGen Solutions (Autodesk Gold Partner)· Operations Manager / Dev Lead
RemoteLed development and operations: Autodesk Fusion 360 plugins, reporting data flows, and Salesforce admin/development. More GCP, more enterprise integration.
- Fusion 360
- Salesforce
- GCP
- Leadership
Webhead Solutions· Software Tailor · Full-Stack Engineer
NowFractional and project-based software for small, growing companies. A trusted ~1.5-year KAEB engagement anchors the work, alongside my own products: OH SCRAP, Solved The Problem, and Sovereign Command Center.
- React
- TypeScript
- Multi-tenant SaaS
- AI
Measured to your business. Cut to fit. Altered as you grow.
Let's build the system that runs your business.
Tell me about the system that hurts. I'll reply with a straight read on what it takes to fix.
Send a message
Goes straight to my inbox. I read every one.