AI · Product · Operations · Technical Systems

I decide what gets built, then write it so someone else can run it.

I build and operationalize products, technical workflows, and operating systems across software, health, research, and early-stage companies. The work starts with an ambiguous problem and ends with something another person can understand and continue.

View selected work Contact
Portrait of Xander Griego

The first 90 days

A new problem is checked at these three points. The results below came from work run this way, in a full-time role or a shorter engagement.

Day 30

I sit with the people who do the work. There is a short map of where it breaks, a ranked list, and the first procedure or install. What we will measure is written down before anything is built. If the fit is wrong, we stop.

Day 60

One change is in use on real work. The people who will keep it have used it without me in the room. Training is a working session and a written guide.

Day 90

The change has an owner, a review date, and a way to tell if it slipped. I can name what I would not build next. You keep the procedure and the decision record.

Results

~90%less rewrite on generated code, after a change had to read the project records on a codebase of about 500,000 lines
~8×active use of the university research systems, from about 20 people to about 150
100+data stores, and about 30 workflows, rebuilt by two people within a week from documents written first
1,500+registered users of NRTR Health, about 1,100 monthly active, and 30 paying partners

What I work on

The problems rarely stay inside one department.

Strategy and operations

Priorities, operating models, measurable workflows, implementation plans, partnerships, pricing, and processes another person can repeat.

Early-stage company operations, customer implementation, research operations, commercialization, contractor coordination, and internal systems.

AI and automation

AI inside a real workflow. The model is not the product.

Context for coding tools, research pipelines, workflow automation, model boundaries, evaluation, a record that stays on the machine, and a stop when the source is missing.

Product and systems

From the problem through architecture, implementation, adoption, and the next revision.

Software products, data moves, research systems, a health platform, specifications, user education, and the operating work around them.

Documentation and enablement

Complicated work written so someone else can do it correctly, and can tell when to stop.

Developer documentation, architecture, specifications, procedures, implementation guides, research files, training, project records, and handoffs.

Work

NRTR Health

Making a code change read the note that describes it

Problem

Coding tools could see the files in the prompt and not the decisions around them. A change could be valid in one file and wrong for the rest of a codebase of about 500,000 lines. The cleanup took weeks.

What I owned

The operating model that ties source files, decisions, and the notes a tool must read. About 1,900 records sit on files and functions. Before an edit, the tool reads the records for those files. If none come back, it stops. After the edit, the records are updated. Release checks covered the build, the screens, and the logs. A person still approved the release.

Result

Rewrite on generated code fell about 90 percent. A comparable development cycle went from about four months to about two weeks. New interns could run the workflow after a few hours.

Read the system · Weekly review

Cycle and rewrite

Rewrite remaining~10%
Four months is drawn as sixteen weeks, so two weeks is one eighth of the bar.

NRTR Health · Nov 2022 to present

Turning scattered health data into a product people can run

Problem

NRTR started from a hard personal health problem. History, observations, training, nutrition, and practitioner review lived in different places. Nothing connected the evidence, the decisions, the uncertainty, and what changed over time. Hosted stores also made cost, security, and vendor dependence a barrier to keeping a free product that clinics could use.

What I owned

Product direction, the architecture, the documents, the research pipelines, the practitioner course, partnerships, pricing, and contractor coordination. The record moved onto the operator’s machine. About 20 outside services are reached through one internal boundary. A model may suggest. The application writes the decision and names the inputs. If the model is down, the screen says the suggestion is unavailable.

Result

Two people rebuilt more than 100 data stores and about 30 workflows within a week, from documents written before the move. Variable cost per active user went from about $2.15 a month to about $0.03 a year. 1,500+ registered users, about 1,100 monthly active, and 30 paying partners, with no paid advertising. A four-week course of about 20 modules. 30+ practitioners. About 89 percent passed the first sitting. Everyone who stayed finished after the extra teaching.

NRTR Health · GitHub · Decision record · How it is divided · One module

Variable cost per active user, per year

$2.15 per month is $25.80 per year. The after bar is drawn wide enough to see. The ratio is about 860 to 1.

Practitioner course

Passed the first sitting89%
Finished after extra teaching100%

University of Colorado Anschutz · Jul 2022 to Jan 2024

Getting study teams onto the systems

Problem

Study teams, regulatory files, and the electronic systems did not share one procedure. The same startup work broke in the same places. About 3,500 research studies were open. I did not manage the investigators.

What I owned

Study-startup coordination, the electronic research systems, the procedures, training, and the files that had to be right before approval. That included OnCore, the regulatory binder, and data capture. I rewrote the steps that kept generating the same tickets.

Result

Active use rose from about 20 people to about 150, including major oncology teams. Repeat support fell about 80 percent. 250+ studies went through setup and handoff. 150+ regulatory files were prepared or checked. 50+ procedures let staff finish the work without a ticket.

When a file is ready

Adoption

250+studies
150+files checked
50+procedures

Sigil42 · Dec 2025 to Sep 2026

Turning customer work into a process someone else can run

Problem

Near-zero implementation, fragmented communication, and scope that kept growing. The risk was a build nobody could staff after the first install.

What I owned

Technical direction, security, the review cadence, specifications, statements of work, sales conversations, and the numbers used to judge whether an install was worth keeping. A contractor received the assigned piece and did not receive the rest of the design. On the consumer chat product I set the offer and the message. Other people own that application’s code.

Result

A codebase of about 200,000 lines reached a working version in about three months. About 30 specifications. Four statements of work. One sales conversation became a formal partnership. The first installs were done in the room, then written as procedures that could be staffed.

One assigned piece

Delivery

~3 moempty build to a working version
6people on the core team
~30specifications
Month 0 Empty implementation. Scope still moving.
During The decision is written before the task is assigned.
Month 3 Working version. One formal partnership.

Experience

NRTR Health / Neuramod

Nov 2022 to present

A health platform spanning software, research, data architecture, documentation, education, partnerships, and how the product is sold. The record stays on the operator’s machine. The same body of work covers research pipelines, practitioner workflows, specifications, pricing, and contractor coordination.

Sigil42

Dec 2025 to Sep 2026

Customer workflows turned into processes another person can run. Technical direction, security, delivery, specifications, sales, partnerships, retention, and the numbers used to judge an install. A core group of about six, plus contractors.

University of Colorado Anschutz Medical Campus

Jul 2022 to Jan 2024

Clinical research operations, study startup, regulatory systems, process improvement, documentation, system adoption, training, and user support for an environment with about 3,500 open research studies.

Independent advising

2017 to 2022

Offers, positioning, partnerships, and operating steps that 20+ owners in health, fitness, and personal development could keep after the engagement. A Nashville event drew 400+ people. Audience channels together passed 10,000 followers.

Writing and docs

A useful document says what the system does, what it does not do, which inputs matter, what success looks like, what can fail, and what to do next.

01

Set up a local NRTR environment

From an empty machine to a client, a service, a database, and one seeded record.

Setup 02

Project records for a code change

About 1,900 records. If no record comes back, the tool stops.

Procedure 03

How the NRTR platform is divided

Who owns the record, who may suggest, and what stays on the machine.

Reference 04

When the local environment fails

Connection refused, empty lists, bad tokens, and a service that exits on startup.

Troubleshooting

Education

Western Governors University

Master of Business Administration

2021

Bachelor of Business Administration and Management

2020

Contact

Open to work in AI, product, operations, technical systems, implementation, program leadership, documentation, and the work that connects them. If the problem crosses departments, needs a fast start, and needs both the build and the follow-through, I want the conversation.

xandergriego5@gmail.com