The engine
erdosP-3

erdosP-3

An optimization engine that holds the real rules your operation runs on — and recalculates a workable plan every time reality moves. You talk to it in plain language. Named after Pál Erdős.

One change, end to end

What happens when a supervisor asks for something

01 You say it

“Move the two people off gate B12 to the delayed turnaround on A4.”

No form, no configuration screen. Typed or spoken, in the screen you are already working in.

02 It becomes constraints
unassign(staff × 2, gate B12) assign(→ turnaround A4) respect: certs, rest, walk time hold: SLA on B12

The sentence is translated into parameters the solver can actually work with.

03 It re-solves
B12
A4
C7
the reassignment
A4 turnaround
back on time
B12 SLA
still met, 6 min margin
Rest rules
no breach

The whole shift is recalculated, and the trade-off is shown before you commit.

Why it is not a chatbot

A chatbot answers. A solver commits.

erdosP-3 is not a chat layer sitting next to a separate optimizer. It is the optimizer. The language model only does one job: turning what you said into constraints. Every plan that comes back has been checked against the full rule set of the operation.

Build principle — both UI and chat

Operations are run by people working through screens every day, so the interface has to be genuinely good. Every screen also has a conversational layer on top. Not chat instead of UI — chat and UI.

The rules it holds
Qualifications
who is certified for pushback, de-icing, loading
Rest and shift rules
breaks, maximum hours, shift ends
Equipment
what is available, where it is, what it fits
Geography
walking and driving time between stands
SLAs
contractual windows per airline and task
Live position
where every crew, vehicle and turnaround actually is, from on-site sensors

These are not values somebody typed into a form. Position, equipment status and task timing come off IoT devices on the apron, so the plan is solved against what is actually happening.

every completed shift feeds back into the next solve

More of what you can ask

Change the plan in words

“The 11:40 is delayed forty minutes — what does that do to the rest of the shift?”

the knock-on effects across later turnarounds are shown before you commit.

“Who's currently free and qualified for pushback?”

a real answer, grounded in live skills and availability data.

“Keep the SLA, but nobody stays past 14:00.”

two competing constraints, resolved into one plan — or a clear statement that it cannot be done.

Check out the solver
See erdosP-3 running against a live schedule.
Open the solver

See the system in action

We will walk you through how the feedback loop applies to your operational environment.