crolens

Category

AI NEWS AGGREGATOR

Year

2024

Role

Full-stack development, ML pipeline, data architecture

/2.5

Croatian news, curated by machines

News aggregator that pulls content from dozens of Croatian sources on its own. Vector search and clustering via Supabase group related articles together, while OpenAI summarizes them for quick reading.

crolens screenshot 1
crolens screenshot 2
crolens screenshot 3

The Problem

Croatian news is fragmented across dozens of small publishers with no unified view. Readers visit multiple sites to stay informed. Stories get duplicated, important context gets lost between outlets, and there's no way to quickly grasp what's actually happening.

The Approach

Built scrapers that collect articles from 30+ Croatian news sources around the clock. Used vector embeddings via Supabase to cluster related stories together, eliminating duplicates. OpenAI generates concise summaries that capture the key facts across all sources covering the same event.

The Outcome

Aggregates 500+ articles daily into clear, grouped clusters. Users get the full picture of any story in seconds instead of visiting multiple sites. The vector search finds connections between stories that human editors miss.

Tech

  • Supabase
  • OpenAI
  • Next.js
  • Vector Search
/2.11

Convinced yet? Let's build something together.

filip@ipsilon.agency