Co-founder · Infrastructure & data
Toprak Necat Gök
Lived this problem first-hand, building eval for a WhatsApp booking agent. Owns the data pipeline, clustering, and the cloud underneath sparr.
Who we are
We built sparr because one number kept hiding the truth.
Before sparr, we were on the other side of this problem. Building the eval stack for a WhatsApp booking agent meant nine prompt versions, more than 150 test scenarios, hundreds of reviewed conversations, and a faithfulness score of 0.78 that never said which conversation broke, or why.
So we built the tool we wanted back then: personas drawn from your real conversations, graded one segment at a time, with the exact chat behind every verdict.
The team
Two founders, one stubborn question: which customers, and why?
Co-founder · Product & design
Tunahan Büyükgebiz
Turns raw conversation data into personas you can read and a report you can act on. Owns the product experience end to end, and builds our iOS apps.
Before sparr, we shipped the 0.78 that lied to us. Read the full story
What we believe
Three rules we won't break.
01
Names, not numbers.
The report says which segments and why, never a single blended score.
02
Your data, your personas.
Personas come from your real conversations, not a generic benchmark.
03
Rehearse, don't gamble.
Test the agent against them before a single customer does.
Get in touch
Talk to a human.
A real reply from a person. Tell us about your agent and the customers you can't yet see.
General
[email protected]Direct to a founder
- Toprak Necat Gök[email protected]
- Tunahan Büyükgebiz[email protected]
Rather see it on your own agent?Book a demo