PRIYANSHURulesInboxesRecordsTicketsInvoicesFound oneKUMAR
PRIYANSHU KUMARPRIYANSHU KUMAR
Priyanshu Kumar

PRIYANSHUKUMAR

I build software that watches what you can’t.

It reads the places nobody has time to check, finds the few things that matter, and hands them to a person with the proof attached.

See the workSee the workWrite to meWrite to me

(About)

I'm Priyanshu, an AI engineer who works alone. I build software that reads the places nobody has time to check: rule pages01, customer inboxes02, rows of records03. It finds the few things that matter, and hands them to a person with the proof attached.

01 Rule pages01 Rule pages02 Customer inboxes02 Customer inboxes03 Records03 Records
Rule pagesCustomer inboxesReview sitesOrder recordsStock listsSupport ticketsInvoicesContractsRule pagesCustomer inboxesReview sitesOrder recordsStock listsSupport ticketsInvoicesContracts
Reads itFinds itKeeps the proofHands it overSays when nothing changedReads it again tomorrowReads itFinds itKeeps the proofHands it overSays when nothing changedReads it again tomorrowReads itFinds itKeeps the proofHands it overSays when nothing changedReads it again tomorrowReads itFinds itKeeps the proofHands it overSays when nothing changedReads it again tomorrow

What I build

One machine, four jobs. The thing being watched changes. The idea does not.

[01]

Rule changes

A rule changed. You do not have to go looking for it.

It reads the pages where rules are posted, every week, and holds each one up against last week's copy.

Arrives withThe old lineThe new lineThe page
An email showing the old line struck through, the new line, the page it came from, and four other pages marked same as last week.
The email that arrives. Example, with the wording made up.
[02]

Customer messages

Someone asked for help. Now someone sees it.

It reads every place customers write to you and pulls out the ones that ask for something.

Arrives withThe messageWhere it was saidWhen
One list of customer questions, each with where it was said and how long it has been waiting, including one marked not sure.
The list that arrives. Example, with the messages made up.
[03]

Records that disagree

Two lists say different things. You see which rows.

It holds two lists up against each other every night and keeps only the rows that do not match.

Arrives withBoth rowsSide by side
Three rows where two systems disagree, both numbers shown, and a line saying the check ran and when.
The rows that arrive. Example, with the data made up.
[04]

Proof, every time

You never have to take its word for it.

Nothing arrives alone. On quiet days it tells you it looked, and found nothing.

Arrives withThe sourceWhat was checkedWhat was not
A morning briefing listing twenty messages it read, four it flagged, and the message number each flag came from.
My own agent, on a made-up set of messages. Each flag carries the number of the message it came from.

Selected cases

(03)
01 · A US healthcare practiceRules nobody had time to readWhat went wrong: It sent too much at first, so people stopped reading.Read the caseRead the case02 · A small shop in the UKCustomers asking in the wrong placesWhat went wrong: It missed people who were upset but polite.Read the caseRead the case03 · A stock business in the USTwo lists that did not agreeWhat went wrong: One night the check did not run, and nobody noticed.Read the caseRead the case

Things you can open

(04)

The cases above are client systems, so they are anonymous. These are mine. The name is on them, or the code is public, and you can go and read the thing itself.

(Contact)

Writeto me.

I usually reply within a day.

What is it about
Or just write: priyanshukumarmaurya2224@gmail.com