[Jugurtha]

I wrote a bit about it in a Twitter thread[0] and repeated here sometimes[1]. The context is custom machine learning products for large enterprise clients in the mid six-figure range that one or two individuals can pull off.

We've been doing this for seven years and fortunate to have been profitable from the start. Now building our machine learning platform[2] to streamline this and solve actual problems we faced in the real world doing real projects that enterprise clients has paid for. Some projects took us longer than what we think their irreducible duration should have been. It is important to always remember so one does not build a product around imaginary workflows, or around workflows that never produced actual paid products. i.e: when you abstract something into a product, it helps to leverage your domain expertise and it helps having shipped paid product in that space. This is a reminder to people who go into something from a different background and think nicer stylesheets/UI will solve things, or assume people in a domain are morons who never heard of "the cloud" or "Kubernetes". Sometimes you may even have to remind your team looking at other products and wanting features: have we actually ever faced that in real life or is this Aikido stuff/Don Quixote battling imaginary adversaries/problems?

Why: at some point in your consultancy, you'll notice that certain patterns emerge and that you can solve certain classes and clusters of problems you faced delivering software itself with software. That product will help you tremendously if you build it.

Happy to answer questions.

- [0]: https://mobile.twitter.com/jugurthahadjar/status/13106682933...

- [1]: https://news.ycombinator.com/item?id=24972611

- [2]: https://iko.ai