What it really takes to make AI behave in production

On stage, AI is magic. It writes sonnets. It paints pictures. It solves the world’s problems in three seconds flat.

Then you bring it back to the office.

You try to install it in a fintech company with 1,000+ people. You introduce it to regulators. To compliance officers. To real customers with real money.

Suddenly, the magic stops. And the engineering begins.

That is what this newsletter is about.

Deriv<ed> is not for the person who wants to hear about the “limitless potential” of the future. It is for the engineer who has to make it work today.

We are not selling you a dream. We are showing you the plumbing.

We are the Applied AI tribe at Deriv. We don’t have the luxury of “moving fast and breaking things.” Because when things break here, customers lose money.

So, we test. We fail. We fix.

In this newsletter, you won’t find hype. You will find the error logs. The architectural arguments. The open-source tools we had to build because the existing ones were just toys.

It isn’t always pretty. But it works.

The fine print

Who is writing this?

Deriv<ed>’s editor and resident writer is Waqas. He’s our Head of Applied AI. He has a PhD from Cambridge, which means he knows how much he doesn’t know. He believes that an AI system is only as good as the business problem it actually solves.

He is joined by the engineers, data scientists, designers, and product managers in the trenches.

Join to get:

• The blueprint: Technical deep-dives on how we put AI into trading, compliance, accounts, HR, and more.

• The post-mortem: An honest look at what broke. (And why).

• The toolbelt: Frameworks and code we use daily.

Our AI philosophy

• AI should deliver genuine business value, not just automate things.

• The only way to get good at applied AI is to ship, measure, iterate, and share what you learn.

Join us

Subscribe so you’ll never miss our technical breakdowns, system architectures, and lessons from building AI in production. Deriv<ed> posts shall remain free. Always.

If you are the kind of engineer who gets bored when things run too smoothly, if you prefer a difficult reality to an easy theory, get in touch. We have plenty of problems left to solve. And we like to solve them with talented engineers who love AI.