Skip to main content

Work · Enterprise applied AI · Platform & transformation leadership

Build systems that outlast the builder.

I build governed enterprise systems that turn local success into reusable capability and durable ownership. At AWS, that meant an applied-AI platform; at Uber, a global operating model spanning 40+ business lines and 80+ countries.

AWS12 live tools · 27+ reusable components · 4,500+ automated tests · 8 independent production builders · 2 ownership transfers
Uber40+ business lines · 80+ countries · eight-person core team · 35→10 average cycle days

Flagship cases

Flagship cases · two different forms of the same work

01 · AWS · 2024–present

From successful local tools to shared production capability.

Useful AI workflows were beginning to reproduce authentication, integrations, deployment, testing, logging, support, and founder dependence. Individual success was creating a shared architectural problem. I redirected the work toward common foundations, explicit production controls, independent builders, and transfer from the start.

Platform architecture

Local success becomes durable only when capability can be reused, governed, shipped, and owned by others.

01 · Local winsProve useful workDomain workflows establish value and reveal repeated technical needs.
02 · FoundationShare the common layerAuthentication, integrations, reusable patterns, and platform services.
03 · ControlMake production explicit4,500+ tests, logging, release practice, observability, support, and review.
04 · BuildersDistribute release capabilityOther builders can extend and independently ship production workflows.
05 · OwnershipTransfer the operating systemChange rights, runbooks, release knowledge, incident duties, and accountable owners.
12live tools
27+reusable components
8independent production builders
2ownership transfers

What I personally owned

Founding architecture, initial implementation, reusable patterns, platform standards, control design, builder enablement, release practice, and transfer design.

The tradeoff

We gave up some local speed and founder control to gain shared reliability, distributed release capability, and systems that could outlast the original builder.

Judgment preserved

A pricing-control workflow returned PASS, FAIL, or UNCERTAIN at the rule level, removing a 24-hour bottleneck while keeping ambiguous cases with specialists.

What remained

Reusable services, explicit operating practices, independent builders, and owners able to run and improve transferred systems.

Evidence boundary: Official title: Business Technical Developer III. “Applied-AI platform leadership” describes functional scope. Portfolio totals are collective; directional capacity estimates are not audited savings.

02 · Uber · 2021–2024

A global operating model, not a software installation.

Agreement work crossed legal, sales, finance, engineering, regional authorities, implementation partners, more than 40 business lines, and more than 80 countries. A software-only intervention would have digitized fragmentation. I led the operating-model redesign that joined decision rights, process, technology, rollout, support, and adoption.

Global operating spine

Standardize the shared path; preserve explicit exception routes where legitimate variation requires them.

Intake
Templates
Decision rights
Approvals
Platform
Rollout
Adoption
40+business lines
80+countries
35→10average cycle days

What I personally owned

Operating-model and transformation leadership, governance design, cross-functional alignment, delivery coordination, adoption, and executive communication.

The tradeoff

We standardized the global spine while preserving explicit exception paths where geography, agreement type, business line, or risk required legitimate variation.

Why the result moved

The 35-to-10-day outcome came from coordinated process, governance, templates, approvals, technology, rollout, and operating discipline—not configuration alone.

What it taught me

Architecture, process, governance, adoption, and ownership are not separate projects. That lesson became foundational to the platform model I later built at AWS.

Evidence boundary: The 35→10-day result is a cross-functional organizational outcome with public corroboration from DocuSign; it is not a claim of sole authorship or software-only causation.

The decision model

BoundaryFoundationControlAdoptionTransfer

Across both cases, the same five decisions recur. The model is less a delivery sequence than a test for whether a system can become trustworthy, reusable, usable, and independent of its original builder.

01

Boundary

What is the decision?

Define evidence, accountable authority, exclusions, uncertainty, and escalation before implementation begins.

02

Foundation

What should be shared?

Build reusable services and workflow patterns instead of accumulating bespoke local wins.

03

Control

What makes it governable?

Authentication, tests, logs, observability, rollback, support, review, and release ownership.

04

Adoption

What makes it usable?

Roles, interfaces, training, feedback, operating burden, and specialist review paths.

05

Transfer

Who owns it afterward?

Access, runbooks, release knowledge, change rights, incident duties, and accountable ownership.

The pattern is consistent: consequential work becomes durable when architecture, governance, adoption, and ownership are designed as one system.

Public systems laboratory

Where I test the method publicly.

The legal and clinical demonstrations apply the same model where provenance, uncertainty, review, and legitimate authority matter more—not less. They are public testbeds for bounded outputs and accountable human judgment, not claims of deployed client impact.

Authoritative sources
Visible uncertainty
Bounded outputs
Accountable review

Status: public demonstrations, not deployed client systems, professional advice, or evidence of clinical or legal efficacy.

WHERE I FIT

When the system matters more than the tool.

I’m most useful when an organization has moved beyond isolated automation and needs technical architecture, workflow design, production governance, adoption, and long-term ownership solved as one operating problem.

For a separate private advisory relationship, read Private Advisory.