Nov 2019 – May 2024 · 4 yrs 6 mos · San Francisco Bay Area · Hybrid

• Product & Strategy Leadership
• Led data science and AI for the Google Nest ecosystem across hardware, software, and subscriptions (Nest Aware, Nest Renew). Partnered with Product, Engineering, and Executive leadership to shape AI strategy, product roadmap, and business decisions presented to the CEO and Board of Directors.
• AI Product Innovation
• Served as Product Manager for Sparkle AI, a patented LLM-powered SQL generator that delivered insights through natural-language chat. Launched Google Home’s first Generative AI feature, enabling conversational intelligence and setting the foundation for AI-powered user experiences across devices.
• Team Leadership & Growth
• Scaled the Data Science organization from 4 to 15 FTEs across the U.S. and India. Led 30+ Data Scientists and Engineers across Google Devices (Nest, Pixel, Fitbit, Watch) to measure ecosystem value through causal modeling and experimentation, directly informing multi-billion-dollar investment priorities.
• Data Infrastructure Modernization
• Modernized Nest’s data infrastructure by migrating legacy systems to Google’s unified tech stack. Delivered new logging, experimentation, and analytics frameworks supporting 10+ major product launches and driving a 35% increase in business growth.
• Organizational Impact & Thought Leadership
• Served on Google’s AI Steering Committee to accelerate GenAI adoption, develop AI education programs, and lead company-wide hackathons and training initiatives. Recognized mentor for Women in Data Science and Google’s mentorship program, fostering talent and diversity in AI leadership.

May 2018 – Oct 2019 · 1 yr 5 mos · San Francisco Bay Area

• Empower prosperity with AI/ML at Intuit for Small Business segment, covering Marketing Intelligence, Smart Products, and Network Graph.
• Marketing Intelligence: Enable personalization and optimization of the Ecosystem Life Cycle journey for prospects and customers through AI/ML.
• Unlock Business Insights on sign up, conversion, churn and upsell in Quickbooks ecosystem, and set up AB testing against data-informed hypotheses.
• Develop ML models for audience segmentation and leverage Explainable AI to personlized marketing campaigns and product experiences.
• Smart Products: We build SMART data products with AI/ML for Quickbooks Ecosystem and integrate models to production with AB testing. e.g:
• Smart Money: Real time prediction of all transactions to be personal or business purposes used for tax deduction in QuickBooks Self Employed.
• Smart Group: Leverage Frequent Pattern Mining (FPM) to group similar trips together for bulk review personal or business trips, saving customer's manual swipes for tax deduction on mileage tracking.
• Smart List: Provide real time personalized recommendation of help articles with LDA topic modeling, Random Forest, and deep learning.
• Network Graph: Collaborated with engineering teams to build and mine a network graph operating at the scale of 1 billion nodes. Explored different algorithms for performing entity resolution at this scale using spark pipelines written in scala. Currently working on developing a framework for representation learning from this network, to be used as input features for different ML models

Mar 2016 – May 2018 · 2 yrs 2 mos

• Currently lead the effort on Learning Analytics supporting enterprise sales, marketing and content strategy. I work with a team of amazing people to build data products to facilitate strategic partnerships and enable customer success with Lynda.com and LinkedIn Learning. Our vision is to create economic opportunity for the global workforce through transformative learning.
• Enterprise Sales: Empower Global Sales Organization (GSO) and cross functional teams with solid data foundation, KPI tracking, machine learning models and end-to-end scalable data solutions in Merlin.
• Marketing: Develop marketing propensity model for customer acquisition, define key-audience segmentation for enterprise demand generation campaigns, build automatic data backend to empower enterprise customer engagement and success.
• Content Strategy: Work with LinkedIn Learning amazing content team and cross-functional teams to build content performance management tools for content engagement tracking and deep dive.

Aug 2014 – Mar 2016 · 1 yr 7 mos

• I'm in the sales intelligence team supporting Global Sales Organization (GSO) by developing end-to-end scalable analytic solutions.
• Projects I've been working on:
• Lynda.com Sales Intelligence - Mining the e-learning and skill data at LinkedIn to empower Lynda B2B sales strategy; Building visualization and scalable tools to help prioritize sales force's work.
• Economic Graph Challenge Mentor - Investigate on human capital leveraging big data at LinkedIn, mentoring 2 PhD students from the Wharton School of Business.
• Churn Guard - One stop shop for churn tracking, prediction and mitigation. Lead a team to build churn risk models, provide churn detailed reasons and recommend churn mitigation stories in Churn Doctor. Currently serves as the product manager for product integration and improvement.
• LTS (LinkedIn Talent Solution) Size of Prize - Revamped the SOP model to cover international regions and staffing segments. SOP has been used in territory planning, net ratio and quota management.
• Amplify - Automatic territory planning with optimization modeling. Lead a team of engineers and sales operation analyst to develop the one-click territory planning web portal.
• Team Tech lead - Organize trainings for team on git, Pig, machine learning, wed dev and spark.

Jun 2014 – Jul 2014 · 1 mo · Stanford, CA

• Ignite is a four-week intensive training to learn:
• Core business skills, such as marketing, operations, strategy, accounting, finance and economics.
• Applied skills, such as negotiation, teamwork, public speaking, leadership, and pitching.
• Approaches to learn product design and prototyping.
• Working in a team to develop a plan for commercializing a new product or a new venture
• Together with my team, we developed a comprehensive business plan on a venture idea "SafetyNet: China's first premium emergency medical monitoring service with mobile capability" and pitched this idea to several VC and Angel Investors at the Ignite final presentation session.
• My role involves marketing research, go-to-market strategy, unit economics and etc.

Jan 2014 – Feb 2014 · 1 mo · Mountain View

• Created dj-cloud.us, a music album recommendation engine based on over 6 million album ratings.
• Merged album data from scraping Amazon, Google/Bing/Youtube APIs, and Stanford SNAP lab.
• Leveraged Hive on top of AWS Hadoop to preprocess data, then applied collaborative filtering with Map Reduce using MRJob to calculate pairwise Pearson and Cosine similarity between albums.
• Calculated top ranked albums, filtered duplicate titles with NLTK, and stored data into MySQL.
• Deployed app with Flask on AWS, designed front end using HTML, Bootstrap CSS and JavaScript.

Oct 2011 – Jul 2014 · 2 yrs 9 mos · Stanford, CA

• Developed a clinical data process pipeline with Perl and C++ to automatically segment brain MRI images, compare features against normal subject database, and generate quantitative reports.
• Integrated the data analysis pipeline into current Stanford hospital clinical workflow and deployed on 8 Stanford MRI scanners. The system has processed over 100 patient cases.
• Evaluated different imaging protocols and test-retest reliability with ANOVA regression model.
• Developed a CT/MRI data integration system for a large-scale clinical trial for stroke patients with Python Flask, MySQL, and Bootstrap CSS.

Aug 2005 – Sep 2011 · 6 yrs 1 mo · Nashville, TN

• Investigated the effect of non-rigid registration algorithms on brain morphometry to detect group differences for subjects with Williams Syndrome and children with math disabilities.
• Built software pipelines integrating MATLAB, C and Bash to process data in parallel utilizing hundreds of nodes in a Linux cluster to detect statistically significant group differences.
• Proposed a Support Vector Machine (SVM) classifier with brain features to classify children with low and normal math performances. The leave-one-out classification rate reached 92.5%.
• Prototyped a unit test in Java for an image processing plugin, deployed at a Hudson server.

Sep 2002 – Jul 2005 · 2 yrs 10 mos · Beijing City, China

• Project: Target Detection in Polarimetric Synthetic Aperture Radar (POL-SAR) Images
• Designed a channel-dependent model to simulate spatially correlated sea clutter in POL-SAR images.
• Proposed an Improved Polarimetric Whitening Filter (IPWF) method that enhanced the detection rate and controlled the false alarm rate on simulated and real data, also better preserve the shape of targets.
• Received research grant award of 20,000 RMB from National Key Lab of Microwave Imaging Technology.