Finding a place to belong

Since August 2026, I have developed and operated a public site for discovering university and inter-university clubs and societies. I built it independently and improved it through feedback from people around me. The service gathers scattered recruitment notices into one place rather than asking students to repeat the same search across communities.

My portfolio records a conservative estimate of several hundred cumulative uses. I present this as a usage estimate, not a measured count of unique users. The live site is linked below so visitors can inspect the service itself.

AI-assisted collection brings notices from university and external communities into one database. Classification organizes clubs and societies so users can discover relevant activities, and summarization turns long notices into information that is easier to scan and compare.

The product decision was to connect information processing to a usable discovery surface. Collecting more notices is only useful when their structure and summaries reduce the work required to find a suitable activity.

I considered how model size and reasoning effort change both output quality and token costs. I established and used a cost-efficient agent strategy so the information-processing workflow could run within a limited budget. The decision is about the whole operating workflow, not simply selecting the largest model for every step.

This work connects product development with the economics of repeated AI processing. Collection, classification, and summarization have to remain useful while being affordable to run. I have not published a quantified cost-saving percentage.

I considered paid visibility for recruiting organizations: a banner near the top of the site or increased exposure when clubs and societies are sorted. These are monetization proposals, not a claim of existing paid placements or earned revenue.

Recruitment is concentrated at the beginning of each semester. By late September, traffic appeared lower. The project exposed a structural problem: a useful seasonal directory does not automatically create year-round demand, and building a durable competitive advantage is difficult.

The next product question is how to remain useful beyond a single recruitment window. That problem affects both the value of paid visibility and the service's long-term ability to attract users.

I identified integration with services such as Everytime and Linkareer as a possible distribution direction. The aim is to bring discovery information closer to places students already use and broaden reach. This is a planned direction; an integration or partnership has not been confirmed.

Building and operating the site taught me to consider collection, AI cost, discovery quality, seasonality, monetization, and distribution as connected product decisions.