The thesis
Both sides of the line most people pick one of.
Most people are strong at process but allergic to tech debt, or strong at systems but ship things an organization can't run. We sit on both sides — we design the workflow, build the tooling, and own the operating cadence that uses it. Increasingly that means wiring AI into the work so a small team punches far above its weight, with disciplined guardrails that keep the irreversible calls human.
The one-line version
Senior systems builder (SFIA ~5, reaching 6) with reference-grade AI-adoption maturity, who pairs deep security instincts with fractional-COO operating discipline.
Proof case
The Cadence license server: zero to live.
A partner needed the specialized FlexNet license server behind an atomic-clock ASIC, on AWS. It looked like "stand up a server"; it was a five-domain integration with silent-failure traps at nearly every step. We took it from an empty account to a hardened, running service — in two days.
Five places the tool would have failed alone
- The license had to outlive the machine. Bound it to a dedicated network identity so the instance is disposable — a design you only make if you know the failure exists.
- The plan would have left it exposed. Replaced public-ports-with-allowlist with a private VPN; zero public surface, isolated from all other prod.
- It failed silently, three ways. A headless-install block, an invisible IPv6 stall, a resource-starved box — none threw a clean error. Hypothesis-and-test on undocumented software.
- A quiet setting would have started a bill. Caught a change that would tip a paid tier — pricing and security held in the same head.
- The vendor relationship was mine. Knowing which questions create risk — the cloud-hosting terms, the renewal timing — is human work.
What shipped
Assessed skill · SFIA
Capability matrix
Rated on SFIA across 7 responsibility levels, each row stamped with when it was last demonstrated. It stays current as work ships.
| Capability | Level | Last demonstrated |
|---|---|---|
| Information security SCTY | 6 | Jul 2026 · Cadence license server |
| Solution architecture ARCH | 5 | Jul 2026 · Cadence license server |
| Systems integration & build SINT | 5 | 2026 · MarketCommand |
| Stakeholder relationship mgmt RLMT | 5 | Jul 2026 · Cadence (Cadence AE + VT + CTO) |
| Emerging technology EMRG | 6→7 | Jul 2026 · MQoS lab appliance — 9-round dual-model plan review |
| Programming / software dev PROG | 4→5 | 2026 · WelcomeMed platform |
| Supplier management SUPP | 4 | 2026 · Cadence AE + MQ supply-chain / procurement system |
| Infrastructure operation ITOP | 4→5 | Jul 2026 · MQoS lab appliance — hardened, containerized host |
| Release & deployment RELM | 4→5 | Jul 2026 · MQoS lab appliance — reviewed, revertable, human-gated cutover |
| Database administration DBAD | 5 | Jul 2026 · MQoS lab appliance — identity-preserving DB migration |
| Network design NTDS | 3→4 | Jul 2026 · Cadence (VPN isolation) |
1 Follow · 2 Assist · 3 Apply · 4 Enable · 5 Ensure/advise · 6 Initiate · 7 Set strategy. Hatched = trending to next level.
AI-adoption · Every.to 8 Levels
How far the leverage goes
A different axis from skill depth, from Every.to's 8 Levels of AI Adoption: how much of the work runs on AI, and how much runs unsupervised. A higher level isn't automatically better — the skill is matching the level to the stakes. Center of gravity: Levels 5–7, with Level 8 demonstrated in bursts. (Every.to's own guide uses our always-on agent as its Level-6 example.)
| Level | What it is | Where we land |
|---|---|---|
| L1 Chatbot | Ask-and-answer; copy-paste in and out. | Foundational — we operate well above this. |
| L2 Copilot | Inline suggestions while you work. | Foundational. |
| L3 Agent | AI completes a scoped task with tools; you review the result. | Foundational. |
| L4 Autopilot | AI runs a whole task end-to-end with light supervision. | Foundational. |
| L5 Workflows | Reusable systems and automations — not one-off prompts. | Solid. ~60 version-controlled skills, n8n automations, a semantic “second brain.” |
| L6 Assistant | An always-on agent that works proactively on your behalf. | Reference-grade. Our “Sage” agent runs on a schedule — Every.to’s guide uses it as its Level-6 example. |
| L7 Multi-agent | Several long-running agents operating at once. | Yes. Parallel agents are our default way of working. |
| L8 Orchestrator | A manager agent dispatching sub-agents to run whole processes. | Demonstrated in bursts — a 9-round author→review→reconcile plan-review loop, run to a numeric bar. A standing overnight worker is next. |
Scale: 1 Chatbot → 8 Orchestrator. Highlighted rows are our operating range; Levels 1–4 are table-stakes we work above.
Every angle
Dimension by dimension
Security · Strongest · SFIA 5
A reflex, not a checklist: mesh-VPN as the actual perimeter, zero-plaintext secrets with hash-verified rotation, row-level security backed by tests, human-approved destructive ops. On Cadence we strengthened the posture past the written plan and caught a cost/security tradeoff the AI missed.
DevOps & Infrastructure · Strong · SFIA 4
Idempotent host provisioning, distroless containers, CI with integration tests on disposable databases, replication + failover, staging→prod behind merge gates. Honest edge: scripted provisioning rather than declarative IaC — an active growth area.
Software Engineering · Senior generalist · 4→5
An 800+ commit healthcare platform with live incident response; documented engine bake-offs and adversarial review before shipping; multiple languages, real tests, operator-grade commit history.
Data & Applied AI · Strong · SFIA 4
Multi-LLM orchestration for extraction, QA, and analysis; pgvector semantic search; structured-output validation so AI never feeds a system as freeform text. Judgment about where AI belongs vs. deterministic code.
Systems Integration · Strong · SFIA 5
The connective-tissue work most builders avoid — ads APIs, telephony, CRM, payments, workflow automation, MCP — stitched into workflows that actually run a business.
Technical Leadership · Evident · SFIA 5
Sets standards, runs adversarial review loops, and does the rarest thing: hands verification to a system while keeping accountability human.
Founder / Operator (COO) · The differentiator · 5→6
Full COO at WelcomeMed and fractional COO via TSG: intake→billing pipelines, structured hiring, device-fleet MDM, versioned runbooks, marketing attribution, and the operating cadences that keep strategy and execution in the same conversation.
Why the guide matters
What the alternatives would have cost
| Scenario | Quality | Failure mode |
|---|---|---|
| Me + AI (what happened) | ~95–100% | None material; secured beyond spec |
| A competent engineer + AI | ~90% | May miss the cost trap or relationship nuance |
| A lower-knowledge person + AI | ~60–75% | Weaker security; can't verify the design → latent trap |
| AI fully autonomous | Blocked | Can't create the account, pass 2FA, hold a relationship |
| Outside specialist / managed | ~95% | Fine, but expensive & slower |
A knowledgeable guide raises the reliability ceiling because they QA the AI. A novice raises the odds of a confident-but-wrong result.
The honest part
Development plan
- Cloud-provider networking (AWS): strong transferable instincts; AWS-specific mechanics are the growth edge → rebuild a service on AWS + pursue the SAA-C03 cert.
- Declarative IaC (Terraform): provisioning is scripted; codify it next.
- Measure the AI itself: tag tasks human vs. agent; publish an automation-rate.
- Persistent orchestration: stand up a standing overnight worker.
Want someone who can build the system and run the operation it lives in?
The Senkungu Group →