Client work · the details stay private
HelloPayments: Paying a sales team without the spreadsheet
Payment sales teams are paid from monthly processor reports, and someone has to split them by hand in Excel.
The big goal was automation: take the manual work out of every month-end. I was the lead engineer. Here is the idea of what I built, not the build itself.
- 6 future products approved to build on my architecture
- 80% of AI questions answered at $0 AI cost
Before and after
- Download each processor's reportpain point
- Clean and match it in Excelpain point
- Work out every splitpain point
- Check it all againpain point
DataSwim
- Reports come in
- AI reads any format
- Every split worked out, with a trail
- A clear report to check
Month-end with Mike
Mike runs a payment sales office: a team of agents who sign up merchants and are paid a share of what those merchants process. The steps above are every month. These are the moments that used to hurt. (An example person; the job is real.)
- A new deal
A new processor signs on
Good news for the business, bad news for Mike: a new file format, a new spreadsheet to build, and weeks of fixing it before the numbers can be trusted.
How it's solvedAn AI reads the new format from the first report. No new code, no new spreadsheet.
- Payday
“My check is wrong”
An agent calls, upset. By hand, Mike digs through a month of spreadsheets to find where one number came from, while the agent waits.
How it's solvedEvery split has a trail back to the report it came from. Mike can show where the number came from, and fix it if it is really wrong.
- Any day
Mike has a question
Which merchants are growing? Why was this month lower? Every answer used to mean another evening in Excel.
How it's solvedMike asks in plain English and gets the answer from the data. Most questions never touch a paid AI model, so they cost nothing.
- The next month-end
A check, not a job
The reports come in, the splits come out, and Mike's month-end is reading one clear report instead of rebuilding a spreadsheet.
How it's solvedMost of the manual work is automated, in the CEO's own words below.
Why this is hard
Every processor speaks its own language
Different files, columns and dates for the same kind of money.
Splits go many levels deep
Merchant, agent, office: each level takes its share, and the rules differ.
Mistakes are paychecks
One wrong number means a real person is paid wrong, and trust is gone.
What I built
Month-end, automated
An engine that reads each month's reports and works out everyone's exact split, with a trail you can check.
New processor, no new code
An AI that reads any processor's report format.
AI that costs almost nothing
A gateway that answers most questions without paying for a big model.
The details stay private, like they will for your project. Ask me on a call how it works.
Lubo is a very knowledgeable engineer. He took the time to understand our business first, then built our entire Python back end and automated most of our manual work with residual files.
- My role
- Senior Data & AI Platform Engineer (lead engineer)
- When
- Mar–Sep 2026
- Built with
- Python · FastAPI · PostgreSQL · Next.js
Want a plan like this for your business? Get your free AI blueprint in about a minute.
