Private by design
Local and private inference, with clear boundaries around where your information goes.
ALL OUT OF BUBBLEGUM LLC
Intelligent systems
for work that
actually has to work.
We connect applied AI, local compute, governed automation, and custom hardware to turn expensive workflows into measurable systems.
SYSTEMS THINKING. HUMAN JUDGMENT.EXPLORE THE SYSTEM
APPLIED AIPRIVATE COMPUTEGOVERNED AUTOMATIONSIMULATION ENGINEERING
01ONE SYSTEMS PRACTICE
Two disciplines. The same engineering instinct.
Design around the work. Understand the constraints. Make every part work together.
SOFTWARE / INTELLIGENCE
PHX Applied AI Systems
Private models. Governed agents. Evidence, observability, and a deliberate path from request to useful work.
Explore applied AIHARDWARE / EXPERIENCE
AOG Simulation Engineering
Custom simulator systems, hardware integration, event deployment, and software built for the cockpit.
Explore simulation02START WITH THE WORK
Buying AI is easy.
Building a reliable system
around it is the hard part.
Subscriptions multiply. Pilots stay disconnected. The impressive demo never quite becomes a dependable workflow. We begin with what happens, how often it happens, what failure costs, and how success can be measured.
03PHX APPLIED AI
Intelligence with its feet on the ground.
A system should show its work, respect its boundaries, and know when to ask a human.
Governed agents
Bounded workflows with explicit permissions, ownership, and human checkpoints.
Evidence you can inspect
Retrieval that keeps the source, context, and reasoning within reach.
The right model for the work
Local, cloud, and hybrid routing with deliberate fallback paths.
Costs in plain sight
Token, cost, and performance observability built into the workflow.
You keep the controls
Human approval before consequential actions. A clear record afterward.
- 01Intake
- 02Evidence
- 03Reasoning
- 04ApprovalHuman checkpoint
- 05Action
- 06Audit
REPORTED WORKLOAD / ARCHITECTURE NOTE
57.95 million tokens.
One architectural lesson.
Agent loops can reread a growing history again and again. A reported local workload shows why context, routing, and observability belong in the design from day one.
Inspect the numbers and their limits57.95M TOKENS
- Input
- 51.5M
- Output
- 6.45M
AOG SIMULATION ENGINEERING
The same systems discipline.
At full scale and full speed.
From the frame to the final feedback loop: simulator design, displays, controls, haptics, networking, and the operational details that make the whole experience work.
From components to a complete systemILLUSTRATIVE SYSTEM CONCEPT / NOT A PROJECT PHOTOGRAPH
IN DEVELOPMENT
The Shell
Pit strategy without fighting the mouse. A developing iRacing pit-assistance concept built around deterministic telemetry and explicit pit plans. The optional AI layer explains and advises.
Explore the design and release gatesTHE CLEAN-BOX INVARIANTCLEAN
or one explicit plan.Simulation only. Live validation pending.
or one explicit plan.Simulation only. Live validation pending.
04HOW WE WORK
A useful result starts with a defined problem.
The value is in the workflow—not the number of models, nodes, or agents inside it.
Understand the work
Map one expensive workflow: frequency, friction, risk, and what success would look like.
Build a focused pilot
Define the boundaries, build the smallest useful system, and keep a human in control.
Measure the result
Compare time, quality, cost, and failure modes against the starting point.
Expand deliberately
Scale what proves useful. Improve what the evidence says needs work.
THE PERSON BEHIND THE SYSTEMS
Built by someone who has had to make complex systems work in the real world.
Darien Fleming brings aircraft systems, avionics, custom hardware, and high-fidelity simulation experience to applied AI. The throughline: verify the signal, control the failure modes, and make the result useful to its operator.
Meet DarienLESS GUESSWORK. MORE USEFUL WORK.
Start with one expensive workflow.
If a process is frequent, costly, repeatable, and measurable, let’s find out what a better system could do.