# Xin Hu > Principal Engineer at Google Location: Los Altos, California, United States Profile: https://flows.cv/xinhu Engineering leader with broad experience in diverse technologies and technical challenges, in service and infrastructure development as well as product engineering. Passionate about building scalable yet resilient systems, finding the most efficient solutions, and solving hard problems in creative ways. Specialties: Consumer Application, Software Architecture, SOA, SaaS, SDLC, Distributed system, Web services, REST, NoSQL/SQL, Design pattern, Java, Scala, Node.js, Objective-C, C++, Play! Framework, Kafka ## Work Experience ### Principal Engineer, Google Search @ Google Jan 2022 – Present | Mountain View, California, United States July 2024 ~ Present Tech Lead for the Search serving infra team, focusing on AI foundation, LLM Eval ecosystem, Search next-gen architecture, and Generative Search experience. * AI Mode for Search: AI mode revolutionizes the information retrieval experience through an LLM-powered chatbot handling complex queries with deep search and personalized context. My contribution focuses on strengthening our Eval infra to improve Eval reliability from 60% to 90%, reduce noise from 30% to 7% while ensuring high fidelity (90+ %), and optimize TPU cost. * GenAI Eval Capabilities: led the team through delivering multiple key GenAI Eval capabilities that underpin LLM iteration velocity and business growth, such as replacing human rating with LLMs, Hybrid Eval, Agent Eval, and Prompt Eval. * Eval Fidelity: established the "Eval Fidelity" charter, creating solutions to ensure our evaluation results are trustworthy by improving data scraping, rating quality, and methodology. *Eval Dev Velocity: spearheaded "Intent-Based Eval Optimization" to accelerate evaluation development. This involved creating standardized configurations to help teams efficiently manage their evaluation goals. June 2022 ~ June 2024 Tech Lead for Google Search & Assistant's data science engineering team, driving data-driven decisions and product direction. Focuses on large-scale evaluations, experimentation, metrics, and strategic analysis. * Generative AI in Search (AI Overview): led multiple work streams across the ecosystem to build new capacities in our analysis stack for measuring and projecting the impact of LLM in search horizontal and verticals. Built strong cross-org partnerships to drive product excellence through experimentation velocity. * Next-generation Assistant Eval (Hermetic Eval): Defined the technical strategy and architecture for a novel evaluation solution, significantly improving fidelity while ensuring performance and privacy. Successfully drove team adoption and cross-org buy-in. ### Principal Staff Software Engineer, Consumer WIT (Women In Tech) Lead @ LinkedIn Jan 2019 – Jan 2022 | San Francisco Bay Area * Tech lead for the LinkedIn Flagship product experience organization, driving the “Flagship as a platform” effort to empower LinkedIn's other lines of business in maximizing impact, by establishing a shared vision, defining engineering principles and best practices, and identifying opportunities for evolving flagship platforms. Currently focus on two main areas: a) Establish and standardize metrics spanning health and resilience, iteration velocity, developer productivity, leverage and adoption to measure success and demonstrate impact for platform investment b) Drive trust as a shared responsibility across engineering organizations to prioritize trustworthy engineering design and outcomes across LinkedIn. * Tech lead for enabling real-time relevance in flagship feed: although widely used by industry competitors, real-time relevance is nascent for LinkedIn’s AI stack. The majority of the recommendation systems rely on offline or near-line computation flows. This led to missed opportunities in capturing members' short-term preferences. To enable real-time relevance in feed, we introduced a real-time feature computation pipeline, enabled real-time candidate selection and ranking to enhance members' feed with contextual and adaptive content. This effort has resulted in a meaningful improvement in feed engagement, such as increased daily unique contributors, engaged feed sessions and revenue growth. * Tech lead for the SCP (Sponsored Content Platformization) project: LMS (LinkedIn Marketing Solutions) aims to help advertisers identify and target those who will benefit the most via Sponsored Content, which has proven to be highly effective in enhancing brand awareness, driving engagement, and generating leads. In partnership with the LMS team, I helped conceptualize the SCP project, created a comprehensive tech strategy and led the project through architecture and tech design phases. SCP significantly improved LinkedIn Ads iteration velocity ### Senior Staff Software Engineer @ LinkedIn Jan 2015 – Jan 2019 * Tech lead and architect for the flagship feed team, which owns the LinkedIn flagship feed experience across desktop and mobile. Drove infrastructure optimization (project High5) that platformize the entire feed ecosystem to improve product iteration velocity. The feed ecosystem spans across feed backend, feed frontend, feed relevance, data science, social action, and downstream viral engagement (e.g push notification). As the architect and overall tech lead for project High5, I partnered with 10+ teams to come up with a strategy plus a set of concrete tech solutions to streamline and accelerate the processes for onboarding new content types to the overall ecosystem. * Bootstrapped the Outlook/LinkedIn integration, with a focus on architecture and design. This initiative aims at bringing people and relationship insights from LinkedIn to the Outlook/O365 suite of products to drive the growth of engaged quality members for LinkedIn/MSFT. * Tech lead for the company wide initiative WWE (world without email) to enable signup and login to LinkedIn with only a phone number, and allowing users to access the platform functionality without needing an email address. The huge complexity of this project comes from the fact that members' email addresses are used across our backend components, as well as product ecosystem. The primary goal of WWE is to deliver values to mobile first users around the world who do not use email as the primary communication channel. It is one of the most important growth initiatives at LinkedIn to capture mobile first users in emerging market. ### Staff Software Engineer @ LinkedIn Jan 2011 – Jan 2014 * Technical Lead for launching LinkedIn android app version 3.0.0, a completely revamped mobile experience with engaging design, more relevant and personalized content, plus improved performance. This release is a huge success as it finally won our flagship android app a 4.2 star review (was 3.7). * Architected, designed and lead the development of mobile application activation framework which is composed of a back-end service, a front-end service and client side libraries for iOS/android. This framework is a critical investment for LinkedIn's multi-app strategy. Today it drives cross app activation for all of LinkedIn's new mobile apps such as Connected, JobSeeker, SlideShare and SalesNavigator. * Created the component SDK which brought hybrid (HTML5/native) mobile offering to LinkedIn's android platform for the first time. Released multiple important features across all supported mobile platforms(iOS/android/mobile web) using component SDK. ### Distinguished Member of Technical Staff @ Motorola Jan 2008 – Jan 2011 As part of the blur team that delivered the MOTOBLUR application framework for android platform, I provided technical leadership in multiple components. I single-handedly built the complete richtext formatting feature for the email client, also created the social messaging app that aggregates a user's social messages from different web portal (Facebook, Linkedin, Twitter, MySpace, etc.). Both two features are shipped on all of Motorola's android devices. ### Principal Software Engineer @ Motorola/Good Technology Jan 2004 – Jan 2008 Good Technology delivers a set of enterprise wireless software solutions using highly secured end-to-end two-way synchronization. I participated in defining and building the client side components that deliver an exceptional user experience to a broad range of mobile devices. I lead an engineering team through initial scoping, system design, prototyping, implementation and the final release for features that allow mega bytes of data such as attachments and SMIME/secure emails to be sent from a mobile device via Good mobile messaging client, by utilizing a high performance transport layer. ### Advisory Software Engineer @ IBM Jan 2001 – Jan 2004 DB2 Everyplace is a relational database runs on multiple mobile devices and embedded platforms. I implemented core features for the database engine that enable enterprise applications and data to be extended securely to mobile devices. ## Education ### Bachelor in Computer Science Peking University ### Master in Computer Science Indiana University Bloomington ## Contact & Social - LinkedIn: https://linkedin.com/in/xinhu2007 - Portfolio: http://www.linkedin.com --- Source: https://flows.cv/xinhu JSON Resume: https://flows.cv/xinhu/resume.json Last updated: 2026-04-12