Building Bazeta: Why a Simple Classifieds Site Still Needs Smart Tech

Описание и применение

Bazeta – быстрый, глобальный классификатор, использующий локальный AI для мгновенной фильтрации и адаптации интерфейса к более чем 100 языкам, снижая нагрузку на сервер и обеспечивая быстрый отклик даже на медленном соединении.

Задача

Create a fast, globally localized classifieds platform that feels instant and relevant across 100+ languages without heavy server load or bloated JavaScript.

Результат

Bazeta achieves a fast, responsive user experience with instant filtering across 100+ languages while keeping server costs low by shifting matching logic to the client.

Рабочий процесс

• 1. Serve lightweight HTML/CSS with optional progressive enhancement. 2. Detect user language and locale on the client. 3. Run a private on‑device AI layer that performs vector comparisons and heuristics to categorize listings and match user intent. 4. Adjust filter weights instantly based on local analysis. 5. Update the view without a round‑trip to the server.
• 6. Use a flat taxonomy with smart defaults that adapt to location and recent activity.
• 7. Render listings with semantic HTML, falling back to no‑JS mode when necessary.

Доказательства

We implemented a lightweight, private on-device AI layer that runs locally in the browser.

Источник

The goal was to keep the initial load time under two seconds on a 3G connection.

Источник

We reduced server load while improving perceived performance.

Источник