Xcelyst.

Xcelyst is a recruitment platform where the core promise is matching: instead of recruiters paging through every applicant, the system surfaces candidates ranked against the role. The product lives or dies on whether recruiters trust what the matching shows them.

Client delivery at Metafic — TypeScript and React on the frontend, Rails on the API. Much of the frontend work in a product like this is dense, data-heavy UI: candidate lists, filters, and detail views that have to stay fast and legible when the dataset gets large.

AI-assisted ranking creates a UI honesty problem: a score with no explanation reads as noise, and recruiters ignore it. The interesting engineering was at the seam between the matching results and the product — presenting ranked candidates in a way recruiters could act on, and keeping the heavy list views responsive against a relational dataset in MySQL that's queried in every direction at once.

A typed React frontend over a Rails API on MySQL. TypeScript pulled real weight here — recruitment data is full of near-identical shapes (candidates, applications, matches, roles), and strict types are what keep a UI over that data from quietly lying.

The platform is live at xcelyst.ai. For me it was the project that made the React-plus-Rails pairing second nature — the same pairing I now lead with on Dallal.