Technology Executive

Twenty-two countries. One operating model.

I build and run the global technology that keeps distributed operations working, in places where "the network is down" means a truck doesn't move and a harvest doesn't ship.

Get in touch LinkedIn

§ 01

Eighteen years, one company, systems engineer to global platform owner.

The arc

For eighteen years I was at Universal Corporation, the world's leading global leaf tobacco supplier. NYSE-listed, more than 20,000 people, operations in more than 30 countries.

I joined as a systems engineer in Mozambique and left as global owner of the company's agronomy field-data platform, running technology across nineteen to twenty-two countries.

2006

Systems Engineer

Mozambique. Servers, networks and site connectivity across a manufacturing operation and remote agricultural sites.

2009

IT Director

Mozambique. A $160M facility and 35 remote sites. Cut telecom costs 60% through VoIP migration.

2013

Regional Head, Digital Transformation

Six African countries. Field-data capture to roughly 1,300 field users.

2014

Senior Director, Technology Strategy & Field Services

Global platform ownership. Twenty-two countries on five continents.

§ 02

Two pieces of work that describe how I operate better than a list of responsibilities would.

What I actually did

Anchor one

From one country to twenty-two

The platform started in the company's Brazil operation. I brought it into Mozambique and proved it in operation, rolled it across five more African countries, and from 2014 took it global.

By the end it ran in twenty-two countries on five continents, in markets with radically different infrastructure conditions. I owned it end to end: the external software firm building it, the Azure data services and MDM infrastructure beneath it, a device fleet of more than 10,000 endpoints including refresh and support, and the capital case in every deployment country.

Five people reported to me directly. Functionally I worked through the head of IT in every country we operated in, roughly 200 IT professionals accountable for anything touching the platform, its infrastructure or the devices.

Anchor two

Rebuilding it without stopping it

By 2020 the platform was ten years old, which is old for software people depend on daily. I initiated and led its full re-specification and ground-up rewrite, deliberately pacing deployment to create engineering capacity, while keeping it running in production across nineteen countries under a tightening investment climate.

Modernizing something people depend on daily, without stopping it, is the hardest version of that job.

§ 03

Figures I can reconstruct and defend.

By the numbers

countries at peak, five continents

200

IT professionals within functional authority

10,000+

managed endpoints across the estate

$30M+

platform program across its fourteen-year life

60%

telecom cost reduction through VoIP migration

years, one company, engineer to global owner

§ 04

The frameworks I actually work by, not the ones that sound good in an interview. Grouped by the job each one does.

How I work

Frame

Cynefin first

Is this Clear, Complicated, Complex or Chaotic? Complex problems get probes, small tests that produce information, before anyone commits to a hard date.

Decompose

MECE structure

Mutually exclusive, collectively exhaustive. If a status report has holes in its logic, the holes are where the programme fails.

Diagnose

The constraint lens

Throughput is set by one bottleneck at a time. MECE tells you every part of the system. This tells you which part is the only one that currently matters, and that effort spent anywhere else is decoration.

Anticipate

The pre-mortem

Before committing, assume it is a year from now and the thing failed publicly. Not "what are the risks", but write the obituary. People will say as history what they will never volunteer as prediction. The top three causes become the probes.

Forecast

The outside view

A date comes from evidence, not from optimism. What actually happened the last five times anyone did something of this class? That distribution is the anchor, and the detailed plan only adjusts it. It is also the strongest shield there is: a base rate cannot be negotiated down.

Transmit

The diplomatic shield

Absorb executive pressure rather than passing it through. Teams doing deep work need someone standing between them and the noise.

Transmit

Forthright diplomacy

Absolute frankness, delivered with tactical tact. Bad news early and plainly is worth more than good news that does not survive contact.

Transmit

Visual logic

A two-by-two beats three paragraphs in a board room. If a decision cannot be drawn, it usually is not understood yet.

Learn

The four-point check

Never more than eight days without asking: did I shield, was I MECE, did I probe, did I calibrate? The fourth is the one that matters most. A review that only audits how I behaved, and never whether I was right, is a comfort ritual.

The first five are how I find out what is true. The next three are how I say it. The last one is how I check whether the first eight are working. I added the diagnose, anticipate and forecast layers in 2026, after auditing my own method and finding it heavy on transmission and light on truth-finding.

§ 05

What came before, and what I am doing now.

Either side of it

Before. Trading-floor support at Credit Suisse First Boston and two Brazilian asset managers, where downtime was measured in money per minute. That grounding in high-availability, high-stakes environments is still how I think about service delivery.

Now. Transit payments across Latin America, where I designed a delivery readiness framework that was adopted as the regional standard and then named the template for the company's UK, Europe and North America operations. I also authored the AI governance standard the team works under.

Thirty-plus years across infrastructure, security, telecom, electronic payments and programme leadership, in three languages, on four continents.

§ 06

The working templates behind two of the frameworks above. Free to take and use.

Things you can download

Two of the frameworks in section 04 only work if they have somewhere to live. A pre-mortem that ends in a meeting note is theatre, and a forecast nobody wrote down cannot be scored later. These are the actual instruments I use, not illustrations of them. Take them, delete my example rows, and put your own work in.

PDF119 KB

Curriculum vitae

Eighteen years at Universal Corporation, systems engineer in Mozambique to global platform owner across twenty-two countries, plus the current transit payments work in Latin America.

Two pages · Updated 2026-08-25 09:29:14

Download CV

XLSX28 KB

Failure Mode Register

The instrument behind the pre-mortem. A facilitation sheet for the thirty minute session, a running register that carries every failure path across every project, and a summary that tracks the only number that matters: how many failures materialised that nobody had probed.

Four sheets · Updated 2026-08-25 08:30:18

Download template

XLSX27 KB

Decision Journal

The instrument behind the outside view. Capture the forecast, the reference class and the confidence before the outcome is known, then score yourself against what you actually said. It builds a reference class library as you go, and it calculates whether you are overconfident, which most people are.

Four sheets · Updated 2026-08-25 08:30:18

Download template

Both workbooks carry a legend, one worked example row from real transit payments work, and live formulas. Nothing in them is a screenshot of a method. If you use one and it tells you something uncomfortable about your own calibration, that is the tool working.

§ 07

Open to senior technology leadership roles, fully remote.

Get in touch

ajgemelli@gmail.com

Download CV

Based

Virginia, United States. US permanent resident, no sponsorship required.

Languages

English, Portuguese, Spanish.

International

Lived in eight countries, worked across more than thirty.

Education

BS Computer Information Systems, Nova Southeastern University.