What I take on
Any AI Problem.
Built by me, start to deploy
If something is slow, expensive or done by hand, bring it to a call and I will tell you whether AI is the answer.
Why this gets built
A lot of what gets sold as an AI problem is a process problem with an API on top. I would rather say that on a call than build the wrong thing carefully. The work I take on runs across fintech and prop trading, healthcare products and operations, and the pattern underneath is the same in all of them: something is done by hand, or paid for per user, and it does not have to be.
What I build
- Half an hour on what you run now, where it is slow, and what it costs you per case.
- A plain answer on whether AI belongs in it. Sometimes the fix is a query, a queue or a form, which is cheaper for you and faster for me to say than to build.
- Where it does belong, a first slice that can be proven in weeks instead of a programme that gets judged next year.
- The three places the money usually hides: the model spend, the infrastructure under it, and any vendor charging per user or per call.
- The in-house version of that vendor, where the volume makes it the cheaper side. The KYC work is the clearest example: ~$3 a verification with a vendor, ~$0.001 built in-house, at 200K+ a day.
How long it takes
The call is half an hour. Anything that follows it gets its own estimate before it starts.
I do not quote before the call. I would rather lose the work than price something I have not seen.
What ships
- A read on whether AI is the right answer
- Where the cost actually sits
- A first slice with a number attached
- A written estimate before anything starts
- An honest no when that is the answer
Tech stack
Next step
Let’s talk, 30 minutesNeed this built?
Tell me what you’re running and where it breaks. Half an hour is usually enough for me to say what it takes and what I’d do first.