How this platform works

The problem this addresses

Young people in Lamu are not short of ability. They are short of information — about what exists, what it asks for, when it closes, and how to get from where they are to where it is. Opportunities are scattered across websites, noticeboards, WhatsApp groups and word of mouth, and by the time news of one arrives it has often closed.

That is an information problem. Afrigender Analytics is a research and data organisation, and information problems are what a research organisation is for.

Where the information comes from

Every opportunity in the directory has a recorded source: the organisation that published it and the web address where it was found. We prefer sources in this order:

1. the official website of the organisation offering the opportunity;
2. an official government source;
3. an official institutional source, such as a university;
4. a verified partner organisation;
5. another reputable source.

The principle behind this is Afrigender's ordinary research standard, applied to a new kind of record: if we cannot trace it, we do not present it as verified.

How an opportunity gets published

Nothing appears in the directory automatically. Each record moves through a workflow that a person controls:

1. Draft

   An editor records the opportunity and its source.

2. Under review

   The details are checked against the official source.

3. Verified

   An editor confirms the record matches its source. The system refuses to mark anything verified that has no traceable source or no named organisation.

4. Published

   The record becomes public, showing when it was verified and where the information came from. Publishing is refused if there is nowhere for an applicant to actually apply, or if the deadline has already passed.

5. Due for review, then expired

   A nightly check moves opportunities out of active results once their deadline passes, and flags any whose verification has grown stale. An expired opportunity is never shown as open.

How matching works

When you complete the pathway questionnaire, the database — not an AI model — selects which opportunities to consider, using what you told us about your education, location, travel, access and interests. Those candidates are then scored against a set of weighted factors: eligibility carries the most weight, then skills, education, interest, location, experience and availability. The weights are configurable and are recorded with every result.

The score decides which of three groups an opportunity falls into — a strong match, a possible match, or a development pathway — and then it disappears. You are shown the group and the reasons, never the number. A score is a description of a correspondence between two records. It is not a measurement of a person, and displaying it as though it were would be dishonest.

Anything you are not eligible for on its stated requirements is always shown as a development pathway, with the requirement stated plainly, no matter how well it scores otherwise. You should be able to see what stands between you and something you want.

What the AI does

AI is one component of this platform, not the point of it. The database decides what is true; the AI only decides how to say it.

Specifically, the AI assistant:

• receives a closed set of facts assembled by the database, and may speak about nothing else;
• is never given a search tool, and is never asked what opportunities exist;
• has its output checked before you see it — a date, a link or an amount that is not in the source data causes the response to be withheld rather than shown;
• is labelled wherever its output appears.

If the AI is unavailable, every feature still works. You get the matcher's own factual reasons instead of a written-up version of them. Nothing here depends on a model being reachable.

What we will not do

• Invent an opportunity, an organisation, a deadline or a contact detail.
• Show an expired opportunity as though it were open.
• Claim a listing is verified when its source cannot be traced.
• Imply government endorsement, or guarantee employment, admission or funding.
• Publish a statistic from a sample too small to support it.
• Convert your information into research data without asking you separately.
• Require a young person to give up personal information merely to look at what exists.

What this platform will become

The most valuable thing here will eventually not be the website. It will be the structured record of what opportunities reach Lamu, which skills young people say they need, and where the gaps between the two lie — evidence that does not currently exist in usable form for this county.

That evidence will be published as Afrigender research, under Afrigender's standards: with its method, its source, its author and its limitations attached. The data dashboard is where it will appear as it becomes sufficient to publish.

How this is paid for

The platform is operated by Afrigender Analytics and funded from its consultancy work. It is free to use, carries no advertising, and does not sell information.