Sep 2022 – Apr 2023 · 7 mos

• Advisors, Investors and Founding members for multiple startups

Dec 2020 – Apr 2022 · 1 yr 4 mos · Mountain View, California, United States

• Head of the Newsbreak local feed recommendation team, which is in charge of serving the local news to all our users (we had 12M DAU for now!). Built a team of 10 engineers + scientists from scratch and have made significant efforts to increase the local content page views for 70%+ over 4 months!!!
• Lead on the following domains:
• (1) Infra:
• refine existing infra & building next-gen recommendation infra;
• (2) Algo:
• applying cutting-edge machine learning & deep learning techniques to improve user experiences & feed quality.
• (3) Data:
• experiment design, validation & performing recommendation funnel analysis.

Jul 2020 – Dec 2020 · 5 mos · San Francisco Bay Area

• Uber-TL in Twitter Ads Ranking for the following components / services:
• (1) Ads Ranking Service: rank & serve ads for Twitter users, including bid optimization, feature hydration, feature extraction, machine learning prediction;
• (2) Ads Pacing Service: budget pacing for Twitter advertisers, which standardizes the way how our advertisers could spend their budgets in an optimal way;
• (3) Scalable Data Fetch: this provides flexibility for the modeling stakeholders to easily add new machine learning features (e.g., user, ad, user-ad cross / counting features) in online feature engineering pipeline.

Apr 2019 – Jul 2020 · 1 yr 3 mos · San Francisco Bay Area

• Tech-Lead for Ads Ranking Service, which is a service for feature hydration, feature extraction, machine learning model prediction, and ads ranking for Twitter Ads.
• We built ranking service from scratch, starting from 2018 Dec, and launched it 100% to production at 2019 July, which is the core service to rank ads based on model prediction, budget pacing, boosting, bid optimization etc.
• As the tech-lead for ads ranking service, I led the team to do various optimizations on top of that, e.g.,
• (1) scalable user & ad data fetches: which provides easy to use ways to onboard user & ad features, which plays important roles for the modeling engineers to iterate ideas on top of that;
• (2) service optimization & continuity: which improved our service over the ads serving funnel to be able to handle the unprecedented traffic, and optimize the service to reduce lots of resources;
• (3) pacing logic optimization: which re-designs the way how we perform budget pacing, and built a service off the request path to further optimize the serving stack, and create opportunites of revenue improvement on top of that.

Dec 2017 – Apr 2019 · 1 yr 4 mos · San Francisco Bay Area

• Building large scale machine learning infrastructure to facilitate processing online ads prediction services. There are a couple of projects I acted as the primary lead:
• (1) Online prediction system optimization: developed several techniques (e.g., Smart Intersection Order, Optimizing Redundant Logic, Maximize Spatial Locality, Efficient Preprocessing, and Efficient Filtering) to optimize the prediction service.
• (2) Ads Prediction Funnel Optimization: re-evaluate the ads-prediction funnel, and later made optimizations on top of that, leading to tens of millions of total revenue growdth for the company.
• (3) Online & Offline Training Infra SpeedUp: Optimize both the online / offline training infra by leveraging the parameter service framework, which led to significant revenue improvement & iteration speed for model ideas.

Oct 2016 – Jan 2017 · 3 mos · NYC, NY, US

• Structured Data Group: http://research.google.com/pubs/DataManagement.html
• Supervisor: Dr. Cong Yu.

May 2016 – Aug 2016 · 3 mos · Redmond, WA, US

• DMX Group: http://research.microsoft.com/en-us/groups/dmx/
• Supervisor: Dr. Yeye He.