How Baidu Mobile Personalizes Content: Insights into Recommendation Systems and Cloud Architecture

Topic 1: Personalization in Baidu Mobile

Baidu Mobile is a super‑app that goes beyond search to cover users' daily lives, facing challenges such as high information‑acquisition cost and diverse user needs. The goal is to make each user's experience unique by delivering the most relevant information and services instantly.

Rich user tags—including demographics, interests, and intents—drive card recommendations, alleviate cold‑start problems, and help operations target specific audiences. Analyses show post‑95 users favor games and animation, while pre‑95 users prefer shopping and food.

The traditional recommendation system relies on similarity measures derived from collaborative filtering, with item‑based and content‑based approaches combined for Baidu Mobile. The system architecture is loosely coupled via APIs, built on Hadoop for large‑scale offline training and real‑time recommendation APIs.

Key challenges include massive user volume and integrating multiple content sources, while advantages stem from Baidu's extensive machine‑learning resources and abundant user data.

Topic 2: Baidu Mobile Cloud Architecture Design and Practice

The cloud side addresses high concurrency, large capacity, flexibility, reliability, and compatibility across many client versions. A multi‑layer solution includes resource aggregation, caching, diverse databases, dynamic UI rendering, account management, multi‑protocol support, and configuration/operation layers.

Performance optimization covers server capacity, request merging, and page caching. Compatibility is achieved through passive client‑side feature detection, code branching, resource ordering, and specialized logic based on client distribution.

Cloud controllability enables dynamic content, events, and configuration delivery without requiring client upgrades, allowing real‑time updates to UI, behavior, and policies.

The overall cloud architecture diagram (shown below) resolves long product lifecycles, uneven traffic spikes, unstable resource links, and complex requirements.

Topic 3: Front‑End Exploration in Baidu Mobile

The front‑end team uses SASS, Zepto, OpenJs, and an EMD module system, providing solutions for templating, local caching, and SPA implementation.

Engineering is driven by a unified FIS build system with custom plugins for dependency analysis, code linting, on‑demand loading, compression, and resource mapping, ultimately publishing static assets to CDN.

Future front‑end directions include hybrid development models, pre‑installed code bundles, and React Native approaches.