Follow us to discover more Baidu tech insights.
This article explains how Baidu Netdisk transitioned from Spark Streaming to a Flink‑based Tiangong real‑time computing engine, detailing the evolution, reasons for choosing Flink, architecture, configuration examples, business use cases, technical challenges, and future platform plans.
This article shares five practical techniques for using Wenxin Code's Zulu and Chat agents—covering context provision, automatic command execution, rule constraints, inline chat, and Git commit shortcuts—to make AI‑generated Java code more reliable, maintainable, and aligned with project standards.
The ERNIE‑4.5‑VL visual‑language model breaks single‑modality limits by delivering breakthrough image, video, and text understanding across more than 100 languages, offering lightweight yet competitive performance against models like Qwen2.5‑VL, supporting 128K context, dual “thinking” modes, and extensive deployment resources.
This article explains the engineering of mobile video post‑processing techniques—brightness and color enhancement using GPU shaders, linear gain, YUV scaling, gamma correction, adaptive saturation, HSV adjustments, and skin‑tone protection—to improve clarity, contrast, and naturalness while maintaining real‑time performance.
FLUX-Lightning, introduced by PaddleMIX, combines phased consistency distillation, adversarial learning, distribution‑matching distillation, and reflow loss to reduce diffusion model inference to just four steps while preserving image quality, and leverages the CINN compiler to achieve over 30% speed gains on A800 GPUs, surpassing existing SOTA acceleration methods.
This article describes how a development team integrated a suite of AI‑driven workflows—including AI‑generated documentation, code scaffolding, rule‑based IDE guidance, MCP server connectivity, and memory banks—to reduce repetitive tasks, accelerate coding, improve code review, and streamline operations across the entire software lifecycle.
This article explains the importance of audio post‑processing in modern player architectures, outlines a modular FFmpeg‑based framework, details core techniques such as bass enhancement and voice clarity, provides algorithmic insights and code snippets, and shows how to integrate these filters into a playback pipeline.
PaddleOCR 3.1 introduces three major upgrades—a multilingual PP‑OCRv5 model supporting 37 languages with over 30% accuracy gain, a PP‑DocTranslation pipeline for high‑quality multi‑language document translation, and MCP server support for flexible AI application integration—accompanied by detailed CLI usage, demo scenarios, and open‑source resources.
This article outlines Baidu’s innovative approach to building its search data platform, detailing the design of wide‑table models, the upgrade to a Spark‑based fusion computation engine, and the new Turing 3.0 service delivery framework, which together deliver higher efficiency, lower cost, and faster, more reliable analytics.
This article explains how Baidu’s next‑generation data platform Turing 3.0 integrates Apache Iceberg to solve the inefficiencies of the legacy MEG stack, detailing ecosystem components, migration strategies from Hive, table‑level optimizations, and future roadmap for high‑frequency, low‑latency analytics.
