Haidian District, Beijing • hi#noahyao.me
Experienced Backend Developer and Systems Architect with over 8 years in backend development and server architectural design. Proven expertise in creating and managing scalable, high-traffic consumer-facing products, social networks, search engines, and B2B applications. Passionate about coding and solving complex problems, bringing a wealth of experience in various technological environments.
Developed and maintained the core product architecture and system infrastructure, including high-performance dual-protocol gateways (HTTP + WebSocket), a scalable user system handling millions of users, a smart asynchronous task scheduling platform, and a real-time, low-latency interactive VR tour product.
Led the design, development, and maintenance of the entire search system for Dongchedi, ByteDance’s automotive platform. Focused on developing query analysis from scratch and extensibly refactoring server APIs to enhance performance and scalability.
Developed and maintained the commercial activity system (voting, lottery, and Qianglou) and native advertising system for Baidu Tieba. Additionally, created a foundational, high-traffic, and SLA-sensitive configuration platform for Baidu APP.
Contributed to the development and performance enhancement of PC-based fundamental services, large-scale data sentiment analysis, and premium features for Baidu Tieba, Baidu’s central social platform.
Designed a Golang framework to facilitate the migration from a PHP core service (handling 200,000 QPS) to Golang, including arranging milestones and backup plans, resulting in saving over 30 instances without incidents. Developed a robust HTTP full-feature gateway capable of sustaining 2 million QPS, and a WebSocket gateway supporting more than 2 million concurrent users.
Led the server-side architecture design and development work for Beike's VR house viewing business, responsible for designing and developing pure memory real-time communication services, supporting VR guided tours business with tens of thousands of people online (peak single-machine value). Through optimization, the link-building performance was improved from 1200ms at PCT80 to 200ms, and an elegant release plan was implemented for WebSocket long connections' special scenarios, extending this technical solution to other long-connection application scenarios.
Accomplished a complete overhaul based on a new architecture involving API+DA+BS+AS, including the splitting and restructuring of DA, BS (using ElasticSearch as the data recall source), and the migration and transformation of the C++ version of the AS module (mainly involving Ranking). This new architecture significantly improved iteration efficiency, ensuring product requirements could iterate on a daily basis, and strategy requirement cycles were shortened to 3 days. Further integrated GO-Jieba, NLP, and other segmentation/rewriting/error correction services, and realized a DA module based on open-source libraries, effectively supporting high-performance intent analysis and segmentation processing in the automotive vertical domain. The new system achieved high performance (PCT99 5ms) and high availability (stability reaching 99.99%+), and can support minute-level interventions for platforms and BOTs, flexibly meeting varied operational demands.
Mobile Baidu App, as Baidu's core mobile product, integrates both recommendation and search functions. In this project, led the design, development, and optimization of the App's basic APIs (such as hot updates, cold updates, etc., with an average QPS of about 5000+, SLA 99.999%). Through the application of a multi-layer storage system (APC/Mysql/BdConf/Redis), successfully restructured the business filtering and issuance mechanism, and upgraded the configuration platform to provide technical support for the disaster recovery system. Additionally, the new configuration platform can handle peak traffic up to 800,000 CPS, process tens of millions of CUID set judgments, implement 13+ kinds of Meta filtering strategies, and optimize traffic through BloomFilter technology.
Responsible for the design, development, and maintenance of Baidu Tieba's native advertising system, achieving a Click-Through Rate (CTR) of 0.28%. This involved building the bidding system, advertiser system, and performance evaluation system, culminating in a comprehensive closed-loop computational advertising platform. Implemented real-time streaming log integration (utilizing nginx, lua, kafka, and storm) to enable the development of subsequent operational and delivery tools. Additionally, led the restructuring and upkeep of Baidu Tieba's commercialization activity system, which included functionalities such as voting, lottery, and building floors. Contributed to a significant commercial project in 2015, composing a thread with over 14 million+ floors, thereby enhancing Tieba's influence in the entertainment and media sectors.
Linux, MacOS, OpenWRT
Vim, Git, Subversion, CTEX, Matlab
Golang, PHP, Shell
MySQL(DDBS), Redis, Object Storage, Memcached, ElasticSearch, CockRoach(NDB)
Kafka, NSQ, Storm