Utah Tech Labs / Applied
See what happens when technology meets the real world.
Go inside the workflows, systems and engineering behind AI that actually does the work.
Demonstrations, technical walkthroughs, product builds and conversations from teams working hands-on across operations, AI, data, software and connected systems.
Explore AppliedFeatured
Start here.
What happens between a construction plan upload and an AI takeoff
Sheet ingestion, scale detection, symbol recognition and the review step estimators actually trust. A run through the pre-construction pipeline end to end.
Open >
Why the data layer comes before the agent
The most common reason an agent pilot stalls is not the model. It is that no system can answer, reliably, what is true right now.
Open >Our collections
Eight ways into the work.
Collections are editorial, not filters. Each one follows a thread through implementation, from the agent doing the work to the infrastructure keeping it up.
Most watched
What operators open first.
How an exception agent handles a shipment before a human gets involved
A late carrier ping, a missed appointment and a customer promise at risk. We follow the agent from signal to resolution, and the point where it stops and asks.
What happens between a construction plan upload and an AI takeoff
Sheet ingestion, scale detection, symbol recognition and the review step estimators actually trust. A run through the pre-construction pipeline end to end.
Why the data layer comes before the agent
The most common reason an agent pilot stalls is not the model. It is that no system can answer, reliably, what is true right now.
What an AI-first dispatch workflow actually looks like
Intake, triage, scheduling and the follow-up, with software making the first move and the dispatcher handling only what deserves attention.
Agents in the operation
Software taking the first pass on work that used to sit in someone's queue.
Data before AI
The layer that decides whether anything above it can be trusted.
On the floor and on site
Manufacturing and construction workflows, shown in the conditions they run in.
From prototype to production
What it takes for a working demonstration to become a system an operation depends on.
Browse all Applied
Find the exact piece.
Filter by topic, industry and format.
Topic
Industry
Format
Collection
Nothing matches that combination yet.
How an exception agent handles a shipment before a human gets involved
A late carrier ping, a missed appointment and a customer promise at risk. We follow the agent from signal to resolution, and the point where it stops and asks.
Open >
What happens between a construction plan upload and an AI takeoff
Sheet ingestion, scale detection, symbol recognition and the review step estimators actually trust. A run through the pre-construction pipeline end to end.
Open >
Why the data layer comes before the agent
The most common reason an agent pilot stalls is not the model. It is that no system can answer, reliably, what is true right now.
Open >
What an AI-first dispatch workflow actually looks like
Intake, triage, scheduling and the follow-up, with software making the first move and the dispatcher handling only what deserves attention.
Open >
How we move a custom model from training to a production API
Dataset versioning, evaluation gates, packaging and the serving path, the part of model work that decides whether it survives contact with production.
Open >
How device telemetry becomes an operational action
From a signal on a device to a work order someone acts on, including the edge decisions that never should reach the cloud.
Open >
Reading a part number on a moving line, and what breaks first
Lighting, capture, model and the traceability record. A computer vision demonstration on the conditions a plant actually has.
Open >
How a large AI-assisted application goes from design to production
The compressed build path: what AI assistance genuinely accelerates, what it does not, and the controls that keep speed from costing you the codebase.
Open >
Deciding what an agent is allowed to do on its own
A practical model for authority, review and audit, and how teams widen an agent's scope without losing the ability to explain a decision.
Open >
Building a custom GPT that carries a role, not a chat window
Scoping the job, wiring the sources, constraining the answers and putting the assistant where the work already happens.
Open >
The cloud architecture behind an AI system that has to stay up
Environments, secrets, failure handling and cost control for AI workloads that are now part of an operation rather than an experiment.
Open >
The first two weeks of a forward deployed engagement
A session on what our engineers do before writing anything: watching the operation, mapping the workflow and finding the decision worth automating.
Open >How an exception agent handles a shipment before a human gets involved
A late carrier ping, a missed appointment and a customer promise at risk. We follow the agent from signal to resolution, and the point where it stops and asks.
What happens between a construction plan upload and an AI takeoff
Sheet ingestion, scale detection, symbol recognition and the review step estimators actually trust. A run through the pre-construction pipeline end to end.
Why the data layer comes before the agent
The most common reason an agent pilot stalls is not the model. It is that no system can answer, reliably, what is true right now.
What an AI-first dispatch workflow actually looks like
Intake, triage, scheduling and the follow-up, with software making the first move and the dispatcher handling only what deserves attention.
How we move a custom model from training to a production API
Dataset versioning, evaluation gates, packaging and the serving path, the part of model work that decides whether it survives contact with production.
How device telemetry becomes an operational action
From a signal on a device to a work order someone acts on, including the edge decisions that never should reach the cloud.
Reading a part number on a moving line, and what breaks first
Lighting, capture, model and the traceability record. A computer vision demonstration on the conditions a plant actually has.
How a large AI-assisted application goes from design to production
The compressed build path: what AI assistance genuinely accelerates, what it does not, and the controls that keep speed from costing you the codebase.
Deciding what an agent is allowed to do on its own
A practical model for authority, review and audit, and how teams widen an agent's scope without losing the ability to explain a decision.
Building a custom GPT that carries a role, not a chat window
Scoping the job, wiring the sources, constraining the answers and putting the assistant where the work already happens.
The cloud architecture behind an AI system that has to stay up
Environments, secrets, failure handling and cost control for AI workloads that are now part of an operation rather than an experiment.
The first two weeks of a forward deployed engagement
A session on what our engineers do before writing anything: watching the operation, mapping the workflow and finding the decision worth automating.
Applied, in your operation
Bring us the workflow you would rather not run by hand.
Everything here started as an implementation. If one of these looks close to your operation, the engineer who built it can walk your team through it.
Show us the workflowSee the same thing on your own systems.
A working session with the engineers who would run the implementation.