Taking an enterprise HR sync platform from prototype to production
The platform was a prototype with made-up dashboard numbers and syncs that failed partway through. In three months I made it ready for large customers, with real data, separate data for each insurer, user roles and a record of every change.
Insurance technology startup · · 4 min read
Context
An early-stage insurance technology startup. Its platform copies employee records from large employers' HR systems into the business systems insurers use, checking and mapping every record along the way.
The founders had a working prototype, but it was not ready for large customers. Closing the distance between demo and real use was the reason to bring me in.
What I did
- Real numbers on the dashboard. Replaced the made-up dashboard numbers with live data from the database, so every demo showed the real state of the platform.
- Servers that stay up during long syncs. Long syncs kept failing on servers built for short requests. I tried several fixes, then moved the app to a server that runs as long as a sync needs.
- Data checks with clear errors. Rejected records used to fail without a visible error. Now each bad record is flagged with what went wrong, and operators can preview the field setup before running a sync.
- Security large customers expect. Added 4 user roles and a record of every change. Each insurer's data is also kept strictly apart from the others.
- A second business system. Added support for a second business system next to the first, with its own field rules and checks.
- Reliable syncs. Syncs now move only the records that changed. I added scheduling and failure alerts, plus a way for operators to roll back a bad batch.
Results
Before
After
Before
Made-up numbers on the dashboard.
After
Live numbers from the database.
Before
Long syncs failing partway through.
After
A server that stays up for the whole sync.
Before
Rejected records with no visible error.
After
Every bad record shown with what went wrong.
Before
No user roles and no record of who changed what.
After
4 user roles and a record of every change.
Before
No separation between insurers' data.
After
Each insurer's data kept strictly apart.
Before
Only one business system supported.
After
Two business systems supported, each with its own field rules.
What comes next
The engagement ended in June 2026. I handed the platform over to the in-house engineering team, along with its ranked roadmap and the written plans behind each item.
Your developers are busy and the product is stuck?
Book a free 25-minute call to see whether working together makes sense.
Prefer writing? Email me at hello@nicolascava.com. I'll reply within one business day.