(In reply to Alexandru Ionescu :alexandrui from comment #2)
And because this is an improving perfherder task, I would also pagination at the top of the page, not only bottom.
It'll be easier to review if you can keep the two tasks in separate commits and possibly two separate pr's.
Have you gotten yourself set up with docker? If not, you'll need to get that set up to work on the backend: https://treeherder.readthedocs.io/installation.html#server-and-full-stack-development
I suggest filing a bug in the Treeherder component to get access to the Treeherder-production-replica RDS instance (read only) so that you will have ample data for testing out API changes locally.
I think what you'll what to do is look into passing a query parameter with the search term from the search input to the API that is called to fetch data, so the top 10 results will be returned based on that filter (we want to keep the existing pagination since new data is called when a user navigates to a new page) . You'll also probably need to change the search input component or create a new one that is used only for the alerts view so that it fetches data, not just update the state.
This is where you'll want to start to familiarize yourself with the pagination code: https://github.com/mozilla/treeherder/blob/master/ui/perfherder/alerts/AlertsView.jsx#L268
This is where the search input component is used: https://github.com/mozilla/treeherder/blob/master/ui/perfherder/alerts/AlertsViewControls.jsx#L54
Adding pagination to the top would probably be the easiest to start with. :)