Civic Tech
Responsible AI Africa is an ecosystem hub for the responsible-AI community across the continent — directories of organisations and AI indexes, funding opportunities, events, jobs, and country-level AI landscapes. I built it on Payload CMS: the content model and collections, an access-control layer, a transactional email outbox, and the Mantine-based public site with on-demand revalidation so editors publish without a redeploy.
Context
Africa's responsible-AI ecosystem — the people, funders, events, initiatives, and country-level policy landscape — was scattered across PDFs, mailing lists, and social posts. Raia centralises it into one searchable, continuously-updated hub that a non-technical editorial team can run themselves, while the public site stays fast and current.
Constraints
Role & Team
Process
01
I defined the Payload collections — directories, funding, events, jobs, insights, international initiatives, countries — and the relationships between them before building the public UI. Getting the content model right first gave the editorial team a coherent system to fill instead of a pile of disconnected pages.
02
The editorial team can't wait for a deploy to publish. I wired on-demand revalidation hooks into the collections so a single Payload edit invalidates exactly the affected pages — content goes live in seconds, with no redeploy and no stale cache.
03
Newsletter and contact flows can't silently drop messages. Rather than calling the email provider and hoping, I modelled an EmailOutbox collection so every transactional send is persisted, inspectable, and retryable — a queue you can audit instead of a side effect you can't.
Outcome
African countries profiled
International AI initiatives tracked
Directory categories of AI actors
Lessons
Building a content platform for a non-technical team taught me that the real product is the editing experience, not just the public site. Modelling the domain carefully up front — clear collections, honest relationships, a glossary everyone agrees on — does more for long-term velocity than any frontend polish. And treating side effects like email as persisted, inspectable state rather than fire-and-forget calls is what makes a content system trustworthy to run.
