Build log
Can an AI ingest 2,798 email threads without inventing one record?
An AI backfilled 2,798 email threads into my structured archive. The number that matters is zero fabricated records, and the boring tripwires that made zero possible.
Independent tests of AI in real work
Anyone can show you what AI did. I show you how I know it’s right.
I use AI on consequential work, then publish the check that separates a convincing answer from a dependable one — what held up, what failed, and where the result stops.
The operating loop
Question → test → proof
Each piece starts with a decision, not a content quota.
01 / Question
Start with a real uncertainty
What would someone need to decide, trust, or do differently?
02 / Test
Try to make it fail
Use source checks, deterministic rules, and an independent challenge pass.
03 / Proof
Publish the receipt
Show the result, method, evidence, limits, and a way to inspect it.
Latest field test
One problem, with the evidence attached
Built to be inspected
The trust layer is part of the work
Public standards
See the exact release loop: the evidence brief, independent checks, human approval, and exact-bundle promotion.
Who is behind it
I’m Ali, a Data Scientist at Uber and a Doctor of Engineering, with eight years of executive reporting, finance automation and applied AI at Uber and Amazon Web Services.