Alex Villarroel
01Selected work

Physical systems // Business operations // Guarded AI

The work, laid out as evidence.

It's the same discipline whether the work is a control panel or an agent fleet. Each piece is laid out the same way: the problem, what I owned, the system, the constraint, the outcome, and the evidence you can inspect.

I show a handful of pieces instead of a full archive. I label status honestly, and anything I can't confirm for publication yet is marked illustrative. I don't publish made-up numbers. Customer figures, names, and drawings only show up where the customer has approved them.

02The exhibits

Six pieces, each laid out the same way.

Industrial project work

Field and project delivery, scope through commissioning.

Completed
Problem
Manufacturing plants and municipal water and wastewater systems need the drawing, the installation, the logic, the network, the schedule, and the people to all agree before startup. Most of the work happens inside live facilities, on retrofits, upgrades, and integrations, where the plant has to keep running and the shutdown window is fixed.
Role
Contractor and project lead across estimating, labor, materials, controls, electrical systems, documentation, and commissioning. Field first: I walk the existing installation down, document what is actually there, coordinate the outage, plan the work, test it properly, and hand the system back clean and working.
System
Field and project delivery carried from scope through installation and commissioning: power distribution, motor control, drives, panels, controls, instrumentation, and industrial networks, comfortable across the full path from incoming power down to the I/O point.
Constraint
Outages, budgets, schedules, safety, legacy systems, and real people set the limits. Available fault current, SCCR, equipment ratings, and coordination are verified with the engineer or the project documents, not guessed. A system is only finished when it holds under real conditions.
Outcome
Selected commissioned projects, with the conditions found and the conditions left both documented. Illustrative until specific figures and customer details are approved for publication.
Evidence
Customer-approved photographs, one-line drawings, and scope excerpts. Illustrative sample only; real artifacts shown where a customer has approved.

Shop Marshal

Contractor operations, estimate to paid.

Early access
Problem
The second shift. Running real contracting work generated recurring administrative load: estimates, crews, documents, change orders, billing, and cash visibility, reconstructed by hand after the field day ended.
Role
Author. I defined the operating model and the single estimate-to-paid record from the operational problems I hit running real projects. Streamline owns the product; I am its author, not its owner.
System
Shop Marshal is a Streamline product, built in the Constellation lab. It's contractor operations software that carries a job from estimate to paid on one connected record. It's in private early access, still in development.
Constraint
A contractor's money and field records cannot tolerate a lost update or a silent mis-book, so the connected record is the source of truth, not a dashboard laid over scattered tools.
Outcome
In private early access, in development. Never described as live or launched.
Evidence
shopmarshal.com, the early-access product page.

Constellation Labs

The architecture behind coordinated AI systems.

Research
Problem
Coordinated AI systems need architecture that can be pressure-tested, not demonstrated once. Memory, review, operating rooms, and authority boundaries have to hold under real load.
Role
Founder.
System
Constellation Labs is Streamline's private research and architecture lab. It's where the architecture behind the products gets built and proven before it ships.
Constraint
An autonomous system must answer one question before it runs: who can stop it. Authority is designed in, not added after.
Outcome
An active research lab. Architecture proven here becomes the estate's technical proof and produces Shop Marshal.

Streamline Automation Solutions

Guarded AI operations for owner-led businesses.

Operating
Problem
Owner-led businesses turn the owner into the human router between every tool, person, and payment. Recurring back-office work does not move unless the owner routes it by hand.
Role
Founder.
System
Streamline installs guarded AI into a business, so the repeat back-office work gets done across the tools the company already uses. The owner still keeps the call on anything with real consequences.
Constraint
The owner keeps authority over consequential actions. Automation carries the work, never the decision.
Outcome
An operating company delivering guarded AI operations.

Orchestrator OS

An open framework for running AI as a system.

Open source
Problem
AI work assembled from loose prompts has no system: no roles, rules, review, or memory that survive a handoff.
Role
Author.
System
Orchestrator OS is an open framework for coordinating AI agents, context, tools, ceremonies, memory, and review as one system, instead of a loose pile of prompts.
Constraint
Open and inspectable. The framework has to be legible to someone who did not write it.
Outcome
Public technical proof, open source.

AI Prompt Sync

A practical public tool from the larger work.

Operating
Problem
Rebuilding prompt and working-directory context by hand every time you open an AI conversation.
Role
Author.
System
AI Prompt Sync is a practical public tool that moves your prompt and working-directory context into an AI conversation, so you don't rebuild the operating state every time.
Constraint
Free and public, on its own desk, backed by a separate entity so a free tool that also takes donations stays isolated.
Outcome
A live public tool, free to use.
03Where to go next

Where to go from here.

Want the thinking behind the work? Read the Field Notes. For the open systems and tools, see the Systems page. For how I got from the field to running AI systems, read the background. Need commercial implementation, contractor operations, or a direct conversation? Take the matching link below.

Read the Field Notes See the Systems Read the background Contact Alex