Leader engineer of AI/ML platform in PayPal. Led the design and development of PayPal’s AI/ML platform, driving multiple initiatives in near real-time feature engineering, model development, and decision systems.

1. Working on a new initiative to optimize model inference stack for PayPal’s risk system, enabling self-service model deployment and significantly improving scalability of inference leveraging GKE.

2. Served as lead engineer in developing the next-generation decision platform for fraud prevention. The platform supports self-serviceable rule authoring; provides Q&A chatbot and rule creation leveraging GenAI and MCP server; ML feature loading and deployment; enables shadow testing on different rule versions with mirrored production traffic.

3. Led the design and implementation of a recommendation system that delivers merchant and product personalization at scale. The system provides workflow based orchestration on candidates retrieval, filtering, model based ranking and re-ranking. It enables data scientists developing and using various ML features including batch, near real-time features for the model serving. It also provides a vector database for embedding based product search.

4. Acted as architect within the AIML platform team, designing and building critical components of the end-to-end machine learning system, including an online/offline feature store and batch/near real-time feature engineering pipelines.

Leader of marketing tech architecture committee. Deeply work with domain team on the design review and help team solve technical challenge. Agile leader of marketing engagement team.

Key achievements:

1. Work as leader of marketing engagement agile team. The focus of the team is to reach to customer with personalized deal through email and push channel, and drive user engagement and active buyer growth. The system deals with 100M+ users and applies science models to do recommendation and targeting based on use activities.

2. Design and lead the development of eBay's multiple channels customer communication (MC3) system. The goal of MC3 is to deliver coherent message to customer via most engaged channel (email, push and so on) at optimized time.

• MC3 provides metadata service and UI tool to manage message definition and payload.

• It provides restful api for triggers to publish/update/delete message and maintains a message store to persistent message

• MC3 Orchestrator handles application logic such as channel orchestration, scheduling, grouping and suppression.

• Internally it uses Kafka queues for streaming to get high throughput.

MC3 is now powering most of eBay's critical push and email messages. The daily throughput is over 500M+.

3. Design and lead the development of Browse 3+ push notification. When user browses an item 3 times or more in the past 7 days, eBay will send a push notification to remind user before item ends. In this system we use Cassandra to store all user’s browsing activities in the past 7 days. The write throughput is about 10K/s. To minimize user impact, it uses MC3’s built-in cap mechanism to limit the delivery count by day or week.

Leader engineer and Architect in eBay mobile platform team. Design and develop back-end service/system to support eBay mobile experience.

Key achievement:

Designed and led the implementation of unified stream service to power eBay new iPad experience. The service is designed in the way that:

1. Simple interface and easy to use

2.Rule based configuration to return different contents based on device

3.Extensible system to support new content in the future.

4. Highly available to serve 100M+ request per day.

Design and implement in-app notification system for eBay mobile and web. Different with push notification, in-app notification is to pull notifications when use opens App.

Leader engineer and Architect in eBay mobile platform team. Designed and developed back-end service/system to support eBay mobile experience.

Leader engineer of eBay search frontend team

Key achievements:

1. Key designer and developer of most of all search services.

Author of FindingService, FindItemServiceNextGen, FindItemService, FindProductService and other key search services. The services are built on top of eBay SOA framework. Based on different use cases (internal or external users), we model the services in a generic or use case oriented way. We try to encapsulate centralized search logic into the service so that consumers can reuse and meanwhile keep the flexibility of service interface to satisfy the different requirements. These services are widely used by search result page, Mobile, vertical (CSA, electric, and Motors) pages and 3rd party users.