Sharing technical practices from Zuoyebang
This article details the step‑by‑step evolution of a multi‑cloud network at Zuoyebang, covering three construction phases, quality‑improvement measures such as fault prevention and rapid recovery, and ongoing operational governance that together deliver a flexible, high‑availability cloud infrastructure.
In this interview, the chief algorithm expert at Zuoyebang discusses how AI technologies such as NLP, speech recognition, large‑model pre‑training, and knowledge‑graph construction are applied to K‑12 education, covering practical challenges, deployment strategies, and future research directions.
This article describes the production challenges of using BERT for large‑scale text classification at Zuoyebang, explores lightweight alternatives such as knowledge distillation, pruning and quantization, and details a teacher‑student‑active‑learning pipeline that trains a TextCNN model to match BERT performance while dramatically reducing GPU consumption and improving throughput.
This article details the operational mindset, stability framework, and transformation journey of the Zuoyebang infrastructure team, covering service lifecycle management, standardization, cloud‑native architecture, multi‑active deployment, incident pre‑plan platforms, traffic scheduling, monitoring, capacity planning, and future directions for SRE service‑orientation.
Zuoyebang replaced its PHP monolith with the Go‑based Zgin framework to overcome microservice, performance, and cloud‑native limitations, delivering a modular, high‑performance backend that supports service governance, efficient logging, and seamless deployment across thousands of services.
The AISummit AI conference highlighted advances in intelligent voice, with experts from ZuoYeBang, ByteDance, Microsoft and others discussing end‑to‑end speech recognition, pronunciation correction, and high‑quality speech synthesis, and exploring how multimodal pre‑trained models will shape the future of voice AI.
This article details how the Zuoyebang front‑end team rebuilt their PC student live‑class client using Electron, introducing a modular, config‑driven architecture, improved logging, BVManager and player modules, and performance and stability optimizations that reduced CPU/memory usage and crash rates while simplifying maintenance.
This report details a series of experiments on Chinese‑English mixed‑language speech recognition, introducing language‑identification loss and language‑model integration to improve acoustic modeling, reduce mixed error rates, and achieve significant gains over a baseline end‑to‑end ASR system.
This article analyzes the architecture and performance bottlenecks of a mobile live‑streaming classroom, defines measurable APM metrics, identifies root causes such as CPU, memory, GPU contention and signaling issues, and presents concrete optimization techniques—including independent processes, containerization, dedicated signaling channels, rendering and thread improvements—that dramatically reduce memory, CPU and frame‑rate problems.
This article explains how the homework‑help platform uses computer‑assisted language learning and neural network models to automatically evaluate spoken poetry, detailing the evaluation dimensions, reliability metrics like Pearson correlation and kappa, data‑driven feature extraction, ONNX deployment, and continuous model improvement through patented automatic data feedback.
