Head of AI · Forward-Deployed AI Engineering · Denver · mrodgersjs@gmail.com · LinkedIn · rodgersintelligence.com

20+ years taking technology into live operations: a Caterpillar fabrication line, five hospital systems, a $6B health-tech business unit, and now my own AI fleet.

Before the repos, the operating record. Programs I have led add up to $35M in combined impact.

• Health systems. Led teletriage across five hospital systems and 14 emergency departments: ED waits from roughly three hours to 15 minutes for 130,000+ patients. Emergency Care Innovation of the Year.
• Teams and ventures. Founded and scaled 83 Tech Harbor: 28 direct reports, $27M in operating budgets, $28M in venture investments.
• Enterprise AI. At Oracle Health, built the internal AI competitive-intelligence capability that processed 150+ market signals a day and was adopted across every market segment.
• Startups. At EmOpti, delivered $1M in first-year revenue and supported a $10M Series A.
• Now. Founded Rodgers Intelligence Group (RIG) in 2025: this fleet, plus client work in governed AI for regulated operations.

_{Figures are from my resume, available on request. Public code and dated CI receipts below are the part you can check yourself.}

One operator. Four machines. Every receipt public.

# Run the front of the estate in one line (READMEs of every pinned system carry the same path):
git clone https://github.com/mrodgersjs-web/mrodgersjs-web.git && (cd mrodgersjs-web && python3 -m unittest discover -s steward -t . -v)

Languages on this profile: Python, TypeScript, JavaScript, Rust, and C++. The fleet that runs my company counts 366 repositories I wrote or own (118 on GitHub under two handles, 34 of them public on this handle, 248 local and in-flight; the census script and method ship at steward/fleet.json), plus 441 vendored research clones I run against but did not write. Where each capability lives in the estate: skills-to-project matrix. 3 public MCP servers ship inside these repositories. Nothing here is a demo directory.

Not a repo list — an infrastructure. The machine-readable inventory is steward/fleet.json, re-censusable with the script beside it.

Dated results from the October 30 audit, each with what it did not prove:

• Moved 66 container images across hosts with rollback data retained and database and service checks. Table counts and live endpoints did not prove row-level parity, so parity stayed open.
• Fixed an MCP process leak of orphaned stdio servers, then tested disconnects and two LAN cycles with 21 tools exposed.
• Qualified inference routes test by test: chat, streaming, tools, continuation, embeddings, vision, context limits. A route that chats but fails tools does not get used for tools.
• Verified 47 application archives on QNAP storage with SHA-256 checks before anything was removed; bounded extraction, so full restoration stayed open.

I set architecture, priorities, and acceptance criteria; AI coding agents help implement and verify. This is internal RIG work, not customer production. Technical case studies and their evidence limits → · Capabilities by role → · The public field record, with per-case not-proven lines →

Also live, also verifiable, not pinned (six pins is the cap): mesh-studio, jake-studio. Same drill: clone, venv, smoke.

_{rigforge in action: an agent reports "BUILD COMPLETE", the signature check rejects the forged packet. Reproduce it with the command above.}

Cross-checked against the requirement language in 108 live Head-of-AI, VP-AI, Forward-Deployed-Engineer and AI-infra postings, October 2026. The percentages are how often that requirement appeared, not how much I rate it.

Python — the working language of the systems above: harnesses, MCP servers, evals, data plumbing, CI, and this page's own stdlib steward.

TypeScript — the delivery surface, and the half of the Forward-Deployed bar my public profile used to hide. Ten TypeScript repositories exist in the estate; ai-user-testing-mcp and rig-os-agency are public and CI-covered, so the claim is checkable instead of asserted.

JavaScript — mrodgersjs-web-teammate, installable with npx, with a green CI run, plus the jev-route-wasm package (JS wrapper, node:test contract suite, browser playground) heading to npm.

Rust — jev-route: the routing policy that governs this fleet, as a standalone dependency-free binary and a WebAssembly build (wasm/, raw four-export ABI, no bindgen, node test suite). Task and candidates in, choice + confidence + fallback verdict out; hard exclusions, scored right-sizing, a confidence floor, deterministic tie-breaks. 14 tests and a 7-case benchmark gate every push.

C++ — latency-probe: header-only C++17 p99/p999 histogram and token-bucket limiter, the primitives you'd write yourself before reaching for a library. Constant-memory log buckets with an upper-bound guarantee (never underestimates), lazy-refill limiter, 32ns/observe benchmark checked in, CI on gcc and clang across Linux and macOS plus an ASan/UBSan job.

Code owns decisions. LLMs assist transformation. Gates decide if it ships.

• Deterministic before agentic. Any rule that can be a function does not get a prompt.
• Every claim names its source, every completion carries a receipt, every migration is reversible.
• Configured, responds, and accepted are three different milestones. I record each separately, and I write down the ones still open.
• Test through the client that will actually use the result, not the API that proves it exists.
• Local-first and private by default. A hosted model is a per-call decision, not a default.

1. 0–2 minutes. Open rigforge and its latest smoke receipt.
2. 2–4 minutes. Select the pinned system closest to the role you are hiring for.
3. 4–7 minutes. Copy its Verify command and run the public smoke path on your own machine.
4. 7–9 minutes. Inspect the linked receipt's state, date and commit SHA.
5. 9–10 minutes. Read the Boundaries line below and decide whether those boundaries matter for your role.

Steward score. 91 of 100 checks green, 4 red, 5 not measured. Measured 2026-11-02.

Rust and JavaScript ride alongside: jev-route ships the routing policy as a dependency-free binary, and the JavaScript claim is the npx surface above.

• U.S. Army Counterintelligence veteran
• B.S. Industrial Engineering, Iowa State University
• Six Sigma Black Belt
• Project Management Professional (PMP)
• Executive Board Member and Chairman, Milwaukee Tech Hub Coalition (2017–2020)

GitHub Achievements: Pull Shark x2, Quickdraw. 141 commits of original work on the public rig-os-agency fork. MIT-licensed public surface, 34 of 34.

I deploy and govern systems other people trained: no distributed serving layer, no training runs. rig-memory-os is the strongest retrieval artifact in the estate and is still private while client identifiers get stripped. Everything public here has a license, a description, and a build or smoke path that CI has actually run. Public repositories here are deliberate release surfaces, and each one that clears its security, licensing and documentation boundary ships. Five shipped across 29 October: the console's CI, a Rust CLI with a wasm build, a C++17 probe, and the six-designer workflow farm.

Internal RIG operating results are not customer production scale or a current uptime report. Where a case study left acceptance work open, it says so.

Hiring a Head of AI, or someone to take agents from demo to production? Email me with the role and the system your team needs built or fixed, or reach me at rodgersintelligence.com.