ALOT: an HR and skills platform

ALOT centralises and automates human resources for companies of every size. It's the largest application in this portfolio, and the reason is the model underneath it: not a set of HR screens, but a competency system that everything else reads from.

ALOT was already at its second version when I came to it, with the competency model built and in use. What the company wanted next was for the platform to do some of the reading: an HR team preparing annual reports was still working through the data by hand to reach conclusions that were already sitting in it.

That turned out to be two jobs rather than one. The visible half is the assistant. The half that made the assistant work was restructuring what sits underneath it — the data model, and the panel's sections, lists and forms — so the same application is legible to a person and to a language model at once. A screen arranged for a human to scan and a record arranged for a model to reason over are not automatically the same shape, and where they differed, the fix belonged in the structure rather than in the prompt.

The competency engine

• A skills dictionary, with rated scales and options defining what each level means.

• Work posts mapped to the skills they require, and users to the skills they hold.

• Self-assessment and interview responses feeding the same model.

What it drives

• Annual planning, per employee, with responses tracked.

• Training plans built from the gap between a role's required skills and the holder's.

• Emerging-talent identification and client propositions drawn from the same skills data.

• Projects and work posts arranged as org hierarchies.

Three audiences, three panels

• An admin panel, an employee panel and an enterprise panel, each seeing its own slice.

• Realtime notifications, role-based permissions, Excel export and scheduled backups.

• An AI assistant integrated into the platform.

Built by my team at Digiton, where I was CTO. I set the architecture, reviewed the work, and owned SEO, performance, responsiveness and the deploy.