Edouard Silvestre
Back to projects

PoC Innovation · Generative AI

EarlyTech

An AI news aggregator that adapts to each reader’s interests.

  • Python
  • Rust
  • pgvector
  • Embeddings
Try it
Period
Oct 2025 – Mar 2026
Team
Team project, with a modest contribution on my side
Code
View the code
sources · scrapingyoucos θpgvectorFor youIARobotiqueStartupsfig. 11 — EarlyTech · the news, sorted for each reader

The project

EarlyTech is an AI news aggregator. What it selects and how it presents it adapt to each user’s interests: multi-source scraping, embeddings, cosine similarity matching with pgvector, and a Rust API.

My contribution to this project was modest.

Try it.

How an article finds its reader

  1. Step 1Python

    Scraping

    Articles collected from arXiv, GitHub, Hugging Face, Medium and Le Monde.

  2. Step 2

    Embeddings

    Each article and each interest becomes a vector.

  3. Step 3pgvector

    Matching

    Cosine similarity links articles to the users they matter to.

  4. Step 4Rust

    API

    A Rust API serves each user their selection.