Annealed
Specs, standards & submittals
Check vendor data and design documents against the requirements, with every finding linked to the page it came from.
FOR EXAMPLE
Submittal review, requirements traceability, and standards checks.
Less manual work.
More engineering.
AI systems that handle the standards review, document checks, and formatting, so your engineers can design the system and run the calculations. Connected to your data and tools, built around how your team works.
DrawingsPlans & revisions
SpecificationsDesign requirements
Vendor dataDatasheets & submittals
AI systemYour data · tools · rules
Purpose-built AI
- Understand the context
- Use the right tools
- Prepare the next step
Review findingsFlagged for your team
Draft analysesEvidence to verify
Technical reportsPrepared for review
- Prepared by
- AI system
- Checked by
- Your engineer
- Release
- After sign-off
01 / WHAT WE BUILD
Start where the
manual work piles up.
Purpose-built systems for specific bottlenecks, not another tool your team has to work around.
Calculations & reports
Turn test data and calculation inputs into draft reports and summaries, with assumptions stated and sources cited.
FOR EXAMPLE
Test-result summaries, calculation package drafts, and recurring status reports.
Drawings & change tracking
Compare revisions, check drawings against a checklist, and trace what a change touches before it reaches review.
FOR EXAMPLE
Revision comparison, drawing checks, and change-impact tracking.
ONE EXAMPLE / SUBMITTAL REVIEW
A clearer first pass.
Not an unchecked decision.
Submittal review is one example: compare a vendor datasheet against the specification, with both sources cited on every line. Visible evidence, clear limits, and human control guide every system we build.
- Cite the specification and the submittal for every line
- Flag ambiguity rather than guess
- Review and approve before anything is returned
Submittal review · Hydronic pumpHUMAN IN THE LOOP
Illustrative example. Select an item to compare the specification with the submittal.
SPEC / SECTION 23 21 23 · PAGE 5
“Motors shall be 460 V, three-phase, 60 Hz.”
SUBMITTAL / PUMP DATASHEET · PAGE 3
Motor: 208 V, three-phase, 60 Hz.
Voltage does not match the specification. Mark for resubmittal, or confirm the available electrical service before accepting an alternate.
02 / HOW WE WORK
Understand it.
Build it. Make it useful.
A small, focused start. Clear boundaries. A working system, not a slide deck.
Find the bottleneck
Walk through the real process, inspect representative inputs, and identify where automation could earn its keep.
Build a bounded pilot
One workflow, defined inputs, and agreed success criteria. Test quality and usefulness with the people doing the work.
Put it into practice
Connect the necessary systems, establish access controls, and plan monitoring, handover, and ongoing support.
03 / WHO’S BUILDING
A mechanical engineer first.
An AI builder by practice.
I’m Ryan Gallagher. I spent five years as a mechanical engineer designing systems for U.S. Navy ships. I know the drawings, specifications, vendor submittals, and review cycles because I worked in them every day.
That’s where I started applying AI to engineering work. Today I build production AI systems end to end, from understanding the workflow to deployment.
- 2020–2025
- Mechanical Engineer
- 2025–now
- AI Engineer
- Education
- BS Mechanical Engineering · BS Physics · MS AI & ML
- Based in
- Delaware County, Pennsylvania
A FEW PRACTICAL QUESTIONS
Before we build.
Do we need to replace our existing tools?
Not necessarily. The starting point is your current workflow and systems. Any pilot will define the integrations it needs rather than assume a full technology overhaul.
What happens when the AI gets something wrong?
That possibility is part of the design. We agree on evaluation criteria, keep source evidence where relevant, and add human review at consequential decision points. The system’s limits should be visible, not hidden.
How do you handle sensitive documents?
We agree on data access, model providers, retention, and deployment requirements before using client data. Don’t send sensitive or confidential documents in an initial inquiry.
What does an initial conversation involve?
A walkthrough of the task, who does it, how frequently it happens, and what makes it difficult. No need for a polished brief. Just a concrete example of the workflow.
LET’S START WITH ONE WORKFLOW
What keeps your team
doing things manually?
Tell me what the task is, who handles it, and where the time goes. We’ll see whether there’s a practical way to improve it.