The work is the portfolio
This studio runs a newsroom that writes, scores, animates and publishes itself nightly — 150 finished things you can press play on right now. Nobody had to take my word for whether the pipeline works.
The other half of the studio
AI that survives legal
Most enterprise AI work stops at the deck. This doesn't. Strategy first, then a system that actually runs — with consent, frequency caps and audit trails enforced in code rather than promised in a policy PDF.
PATENTS 20PRIOR COCA-COLAENDPOINTS 15,000+IN PROD MEIJERBASED ATLANTA
01
Why this desk and not a consultancy
Twenty patents, not twenty case studies
Across AI, security and IoT. These are the ones with public numbers — granted, examined, and readable in full.
- US 11432994 — Intelligence Engine
- US 11600383 — Networked Theft-Prevention System
- US 2020/0334769 — Dynamic Signage for Electronic Menu Board
I have had your job
Global Group Director of Digital Innovation at The Coca-Cola Company — conversational AI in vending infrastructure, deployed across 15,000+ locations. I know what it costs to get a thing through security review, procurement and brand — and how much of the estimate that is.
02
What you would actually be buying
01AI strategy & architecture
02Governed AI execution
03Retail marketing automation
04MCP tool rails & integrations
05Compliance-as-code
06Durable workflow orchestration
Enterprise-grade AI, governed and shipped. Strategy first, then working, compliant systems — the bridge between corporate infrastructure and consumer magic.
03
CASE STUDY · MEIJER · RETAIL
MOS — Marketing OS
Marketing intent in, governed execution out.
Meijer's marketing automation, rebuilt as a platform. A marketer authors a journey, campaign, A/B experiment, or cohort as one declarative document — no code, no redeploy. It compiles at PR time, dispatches by event, and executes durably against the live opt-out registry. Consent and frequency caps aren't a policy PDF; they're enforced structurally, in code, pinned by ~1,500 tests.
43
215
10
16
~1,500
1
MOSLTemporalKubernetesMCPAzure APIMmPerksEmarsys
04
How an engagement actually runs
A call
Thirty minutes. You describe the mess. I tell you if I'm the wrong person — that happens, and it's cheaper to find out now.
A written read
What I think is true, what I'd build, what it costs, what could go wrong. Yours to keep whether or not we continue.
Something running
Not a pilot deck. A working slice against your real systems, behind your real auth.
Hands off
It ships with tests, docs and your team able to run it. I'd rather you not need me next quarter.
If your AI project has been six months from launch for a year
That is usually not a model problem. It is consent, data lineage, or nobody being allowed to own the decision. All three are fixable, and none of them get fixed by another vendor bake-off.
Tell me what is stuck